INSTRUCTOR’S
SOLUTION MANUAL
KEYING YE AND SHARON MYERS
for
PROBABILITY & STATISTICS
FOR ENGINEERS & SCIENTISTS
EIGHTH EDITION
WALPOLE, MYERS, MYERS, YE
Contents
1 Introduction to Statistics and Data Analysis 1
2 Probability 11
3 Random Variables and Probability Distributions 29
4 Mathematical Expectation 45
5 Some Discrete Probability Distributions 59
6 Some Continuous Probability Distributions 71
7 Functions of Random Variables 85
8 Fundamental Sampling Distributions and Data Descriptions 91
9 One- and Two-Sample Estimation Problems 103
10 One- and Two-Sample Tests of Hypotheses 121
11 Simple Linear Regression and Correlation 149
12 Multiple Linear Regression and Certain Nonlinear Regression Models 171
13 One-Factor Experiments: General 185
14 Factorial Experiments (Two or More Factors) 213
15 2k Factorial Experiments and Fractions 237
16 Nonparametric Statistics 257
iii
iv CONTENTS
17 Statistical Quality Control 273
18 Bayesian Statistics 277
Chapter 1
Introduction to Statistics and Data
Analysis
1.1 (a) 15.
(b) ¯x = 1
15 (3.4 + 2.5 + 4.8 + · · · + 4.8) = 3.787.
(c) Sample median is the 8th value, after the data is sorted from smallest to largest:
3.6.
(d) A dot plot is shown below.
2.5 3.0 3.5 4.0 4.5 5.0 5.5
(e) After trimming total 40% of the data (20% highest and 20% lowest), the data
becomes:
2.9 3.0 3.3 3.4 3.6
3.7 4.0 4.4 4.8
So. the trimmed mean is
¯xtr20 =
1
9
(2.9 + 3.0 + · · · + 4.8) = 3.678.
1.2 (a) Mean=20.768 and Median=20.610.
(b) ¯xtr10 = 20.743.
(c) A dot plot is shown below.
18 19 20 21 22 23
1
2 Chapter 1 Introduction to Statistics and Data Analysis
1.3 (a) A dot plot is shown below.
200 205 210 215 220 225 230
In the figure, “×” represents the “No aging” group and “◦” represents the “Aging”
group.
(b) Yes; tensile strength is greatly reduced due to the aging process.
(c) MeanAging = 209.90, and MeanNo aging = 222.10.
(d) MedianAging = 210.00, and MedianNo aging = 221.50. The means and medians for
each group are similar to each other.
1.4 (a) ¯XA = 7.950 and ˜XA = 8.250;
¯X
B = 10.260 and ˜XB = 10.150.
(b) A dot plot is shown below.
6.5 7.5 8.5 9.5 10.5 11.5
In the figure, “×” represents company A and “◦” represents company B. The
steel rods made by company B show more flexibility.
1.5 (a) A dot plot is shown below.
−10 0 10 20 30 40
In the figure, “×” represents the control group and “◦” represents the treatment
group.
(b) ¯XControl = 5.60, ˜XControl = 5.00, and ¯Xtr(10);Control = 5.13;
¯X
Treatment = 7.60, ˜XTreatment = 4.50, and ¯Xtr(10);Treatment = 5.63.
(c) The difference of the means is 2.0 and the differences of the medians and the
trimmed means are 0.5, which are much smaller. The possible cause of this might
be due to the extreme values (outliers) in the samples, especially the value of 37.
1.6 (a) A dot plot is shown below.
1.95 2.05 2.15 2.25 2.35 2.45 2.55
In the figure, “×” represents the 20◦C group and “◦” represents the 45◦C group.
(b) ¯X20◦C = 2.1075, and ¯X45◦C = 2.2350.
(c) Based on the plot, it seems that high temperature yields more high values of
tensile strength, along with a few low values of tensile strength. Overall, the
temperature does have an influence on the tensile strength.
Solutions for Exercises in Chapter 1 3
(d) It also seems that the variation of the tensile strength gets larger when the cure
temperature is increased.
1.7 s2 = 1
15−1 [(3.4−3.787)2+(2.5−3.787)2+(4.8−3.787)2+· · ·+(4.8−3.787)2] = 0.94284;
s = √s2 = √0.9428 = 0.971.
1.8 s2 = 1
20−1 [(18.71 − 20.768)2 + (21.41 − 20.768)2 + · · · + (21.12 − 20.768)2] = 2.5345;
s = √2.5345 = 1.592.
1.9 s2
No Aging = 1
10−1 [(227 − 222.10)2 + (222 − 222.10)2 + · · · + (221 − 222.10)2] = 42.12;
sNo Aging = √42.12 = 6.49.
s2
Aging = 1
10−1 [(219 − 209.90)2 + (214 − 209.90)2 + · · · + (205 − 209.90)2] = 23.62;
sAging = √23.62 = 4.86.
1.10 For company A: s2
A = 1.2078 and sA = √1.2078 = 1.099.
For company B: s2
B = 0.3249 and sB = √0.3249 = 0.570.
1.11 For the control group: s2
Control = 69.39 and sControl = 8.33.
For the treatment group: s2
Treatment = 128.14 and sTreatment = 11.32.
1.12 For the cure temperature at 20◦C: s2
20◦C = 0.005 and s20◦C = 0.071.
For the cure temperature at 45◦C: s2
45◦C = 0.0413 and s45◦C = 0.2032.
The variation of the tensile strength is influenced by the increase of cure temperature.
1.13 (a) Mean = ¯X = 124.3 and median = ˜X = 120;
(b) 175 is an extreme observation.
1.14 (a) Mean = ¯X = 570.5 and median = ˜X = 571;
(b) Variance = s2 = 10; standard deviation= s = 3.162; range=10;
(c) Variation of the diameters seems too big.
1.15 Yes. The value 0.03125 is actually a P-value and a small value of this quantity means
that the outcome (i.e., HHHHH) is very unlikely to happen with a fair coin.
1.16 The term on the left side can be manipulated to
Xn
i=1
xi − n¯x =
Xn
i=1
xi −
Xn
i=1
xi = 0,
which is the term on the right side.
1.17 (a) ¯Xsmokers = 43.70 and ¯Xnonsmokers = 30.32;
(b) ssmokers = 16.93 and snonsmokers = 7.13;
4 Chapter 1 Introduction to Statistics and Data Analysis
(c) A dot plot is shown below.
10 20 30 40 50 60 70
In the figure, “×” represents the nonsmoker group and “◦” represents the smoker
group.
(d) Smokers appear to take longer time to fall asleep and the time to fall asleep for
smoker group is more variable.
1.18 (a) A stem-and-leaf plot is shown below.
Stem Leaf Frequency
1 057 3
2 35 2
3 246 3
4 1138 4
5 22457 5
6 00123445779 11
7 01244456678899 14
8 00011223445589 14
9 0258 4
(b) The following is the relative frequency distribution table.
Relative Frequency Distribution of Grades
Class Interval Class Midpoint Frequency, f Relative Frequency
10 − 19
20 − 29
30 − 39
40 − 49
50 − 59
60 − 69
70 − 79
80 − 89
90 − 99
14.5
24.5
34.5
44.5
54.5
64.5
74.5
84.5
94.5
3
2
3
4
5
11
14
14
4
0.05
0.03
0.05
0.07
0.08
0.18
0.23
0.23
0.07
(c) A histogram plot is given below.
14.5 24.5 34.5 44.5 54.5 64.5 74.5 84.5 94.5
Final Exam Grades
Relative Frequency
Solutions for Exercises in Chapter 1 5
The distribution skews to the left.
(d) ¯X = 65.48, ˜X = 71.50 and s = 21.13.
1.19 (a) A stem-and-leaf plot is shown below.
Stem Leaf Frequency
0 22233457 8
1 023558 6
2 035 3
3 03 2
4 057 3
5 0569 4
6 0005 4
(b) The following is the relative frequency distribution table.
Relative Frequency Distribution of Years
Class Interval Class Midpoint Frequency, f Relative Frequency
0.0 − 0.9
1.0 − 1.9
2.0 − 2.9
3.0 − 3.9
4.0 − 4.9
5.0 − 5.9
6.0 − 6.9
0.45
1.45
2.45
3.45
4.45
5.45
6.45
8
6
3
2
3
4
4
0.267
0.200
0.100
0.067
0.100
0.133
0.133
(c) ¯X = 2.797, s = 2.227 and Sample range is 6.5 − 0.2 = 6.3.
1.20 (a) A stem-and-leaf plot is shown next.
Stem Leaf Frequency
0* 34 2
0 56667777777889999 17
1* 0000001223333344 16
1 5566788899 10
2* 034 3
2 7 1
3* 2 1
(b) The relative frequency distribution table is shown next.
6 Chapter 1 Introduction to Statistics and Data Analysis
Relative Frequency Distribution of Fruit Fly Lives
Class Interval Class Midpoint Frequency, f Relative Frequency
0 − 4
5 − 9
10 − 14
15 − 19
20 − 24
25 − 29
30 − 34
2
7
12
17
22
27
32
2
17
16
10
3
1
1
0.04
0.34
0.32
0.20
0.06
0.02
0.02
(c) A histogram plot is shown next.
2 7 12 17 22 27 32
Fruit fly lives (seconds)
Relative Frequency
(d) ˜X = 10.50.
1.21 (a) ¯X = 1.7743 and ˜X = 1.7700;
(b) s = 0.3905.
1.22 (a) ¯X = 6.7261 and ˜X = 0.0536.
(b) A histogram plot is shown next.
6.62 6.66 6.7 6.74 6.78 6.82
Relative Frequency Histogram for Diameter
(c) The data appear to be skewed to the left.
1.23 (a) A dot plot is shown next.
0 100 200 300 400 500 600 700 800 900 1000
160.15 395.10
(b) ¯X1980 = 395.1 and ¯X1990 = 160.2.
Solutions for Exercises in Chapter 1 7
(c) The sample mean for 1980 is over twice as large as that of 1990. The variability
for 1990 decreased also as seen by looking at the picture in (a). The gap represents
an increase of over 400 ppm. It appears from the data that hydrocarbon emissions
decreased considerably between 1980 and 1990 and that the extreme large emission
(over 500 ppm) were no longer in evidence.
1.24 (a) ¯X = 2.8973 and s = 0.5415.
(b) A histogram plot is shown next.
1.8 2.1 2.4 2.7 3 3.3 3.6 3.9
Salaries
Relative Frequency
(c) Use the double-stem-and-leaf plot, we have the following.
Stem Leaf Frequency
1 (84) 1
2* (05)(10)(14)(37)(44)(45) 6
2 (52)(52)(67)(68)(71)(75)(77)(83)(89)(91)(99) 11
3* (10)(13)(14)(22)(36)(37) 6
3 (51)(54)(57)(71)(79)(85) 6
1.25 (a) ¯X = 33.31;
(b) ˜X = 26.35;
(c) A histogram plot is shown next.
10 20 30 40 50 60 70 80 90
Percentage of the families
Relative Frequency
8 Chapter 1 Introduction to Statistics and Data Analysis
(d) ¯Xtr(10) = 30.97. This trimmed mean is in the middle of the mean and median
using the full amount of data. Due to the skewness of the data to the right (see
plot in (c)), it is common to use trimmed data to have a more robust result.
1.26 If a model using the function of percent of families to predict staff salaries, it is likely
that the model would be wrong due to several extreme values of the data. Actually if
a scatter plot of these two data sets is made, it is easy to see that some outlier would
influence the trend.
1.27 (a) The averages of the wear are plotted here.
700 800 900 1000 1100 1200 1300
250 300 350
load
wear
(b) When the load value increases, the wear value also increases. It does show certain
relationship.
(c) A plot of wears is shown next.
700 800 900 1000 1100 1200 1300
100 300 500 700
load
wear
(d) The relationship between load and wear in (c) is not as strong as the case in (a),
especially for the load at 1300. One reason is that there is an extreme value (750)
which influence the mean value at the load 1300.
1.28 (a) A dot plot is shown next.
71.45 71.65 71.85 72.05 72.25 72.45 72.65 72.85
High Low
In the figure, “×” represents the low-injection-velocity group and “◦” represents
the high-injection-velocity group.
Solutions for Exercises in Chapter 1 9
(b) It appears that shrinkage values for the low-injection-velocity group is higher than
those for the high-injection-velocity group. Also, the variation of the shrinkage
is a little larger for the low injection velocity than that for the high injection
velocity.
1.29 (a) A dot plot is shown next.
76 79 82 85 88 91 94
Low High
In the figure, “×” represents the low-injection-velocity group and “◦” represents
the high-injection-velocity group.
(b) In this time, the shrinkage values are much higher for the high-injection-velocity
group than those for the low-injection-velocity group. Also, the variation for the
former group is much higher as well.
(c) Since the shrinkage effects change in different direction between low mode temperature
and high mold temperature, the apparent interactions between the mold
temperature and injection velocity are significant.
1.30 An interaction plot is shown next.
Low high
injection velocity
low mold temp
high mold temp
mean shrinkage value
It is quite obvious to find the interaction between the two variables. Since in this experimental
data, those two variables can be controlled each at two levels, the interaction
can be investigated. However, if the data are from an observational studies, in which
the variable values cannot be controlled, it would be difficult to study the interactions
among these variables.
Chapter 2
Probability
2.1 (a) S = {8, 16, 24, 32, 40, 48}.
(b) For x2 + 4x − 5 = (x + 5)(x − 1) = 0, the only solutions are x = −5 and x = 1.
S = {−5, 1}.
(c) S = {T,HT,HHT,HHH}.
(d) S = {N. America, S. America, Europe,Asia,Africa,Australia,Antarctica}.
(e) Solving 2x − 4 ≥ 0 gives x ≥ 2. Since we must also have x < 1, it follows that
S = φ.
2.2 S = {(x, y) | x2 + y2 < 9; x ≥ 0, y ≥ 0}.
2.3 (a) A = {1, 3}.
(b) B = {1, 2, 3, 4, 5, 6}.
(c) C = {x | x2 − 4x + 3 = 0} = {x | (x − 1)(x − 3) = 0} = {1, 3}.
(d) D = {0, 1, 2, 3, 4, 5, 6}. Clearly, A = C.
2.4 (a) S = {(1, 1), (1, 2), (1, 3), (1, 4), (1, 5), (1, 6), (2, 1), (2, 2), (2, 3), (2, 4), (2, 5), (2, 6),
(3, 1), (3, 2), (3, 3), (3, 4), (3, 5), (3, 6), (4, 1), (4, 2), (4, 3), (4, 4), (4, 5), (4, 6),
(5, 1), (5, 2), (5, 3), (5, 4), (5, 5), (5, 6), (6, 1), (6, 2), (6, 3), (6, 4), (6, 5), (6, 6)}.
(b) S = {(x, y) | 1 ≤ x, y ≤ 6}.
2.5 S = {1HH, 1HT, 1TH, 1TT, 2H, 2T, 3HH, 3HT, 3TH, 3TT, 4H, 4T, 5HH, 5HT, 5TH,
5TT, 6H, 6T}.
2.6 S = {A1A2,A1A3,A1A4,A2A3,A2A4,A3A4}.
2.7 S1 = {MMMM,MMMF,MMFM,MFMM, FMMM,MMFF,MFMF,MFFM,
FMFM, FFMM, FMMF,MFFF, FMFF, FFMF, FFFM, FFFF}.
S2 = {0, 1, 2, 3, 4}.
2.8 (a) A = {(3, 6), (4, 5), (4, 6), (5, 4), (5, 5), (5, 6), (6, 3), (6, 4), (6, 5), (6, 6)}.
11
12 Chapter 2 Probability
(b) B = {(1, 2), (2, 2), (3, 2), (4, 2), (5, 2), (6, 2), (2, 1), (2, 3), (2, 4),
(2, 5), (2, 6)}.
(c) C = {(5, 1), (5, 2), (5, 3), (5, 4), (5, 5), (5, 6), (6, 1), (6, 2), (6, 3), (6, 4), (6, 5), (6, 6)}.
(d) A ∩ C = {(5, 4), (5, 5), (5, 6), (6, 3), (6, 4), (6, 5), (6, 6)}.
(e) A ∩ B = φ.
(f) B ∩ C = {(5, 2), (6, 2)}.
(g) A Venn diagram is shown next.
A
A C
B
B C
C
S
Ç
Ç
2.9 (a) A = {1HH, 1HT, 1TH, 1TT, 2H, 2T}.
(b) B = {1TT, 3TT, 5TT}.
(c) A′ = {3HH, 3HT, 3TH, 3TT, 4H, 4T, 5HH, 5HT, 5TH, 5TT, 6H, 6T}.
(d) A′ ∩ B = {3TT, 5TT}.
(e) A ∪ B = {1HH, 1HT, 1TH, 1TT, 2H, 2T, 3TT, 5TT}.
2.10 (a) S = {FFF, FFN, FNF,NFF, FNN,NFN,NNF,NNN}.
(b) E = {FFF, FFN, FNF,NFF}.
(c) The second river was safe for fishing.
2.11 (a) S = {M1M2,M1F1,M1F2,M2M1,M2F1,M2F2, F1M1, F1M2, F1F2, F2M1, F2M2,
F2F1}.
(b) A = {M1M2,M1F1,M1F2,M2M1,M2F1,M2F2}.
(c) B = {M1F1,M1F2,M2F1,M2F2, F1M1, F1M2, F2M1, F2M2}.
(d) C = {F1F2, F2F1}.
(e) A ∩ B = {M1F1,M1F2,M2F1,M2F2}.
(f) A ∪ C = {M1M2,M1F1,M1F2,M2M1,M2F1,M2F2, F1F2, F2F1}.
Solutions for Exercises in Chapter 2 13
(g)
A
A B
B
C
S
Ç
2.12 (a) S = {ZY F, ZNF,WY F,WNF, SY F, SNF, ZYM}.
(b) A ∪ B = {ZY F, ZNF,WY F,WNF, SY F, SNF} = A.
(c) A ∩ B = {WY F, SY F}.
2.13 A Venn diagram is shown next.
S
P
F
S
2.14 (a) A ∪ C = {0, 2, 3, 4, 5, 6, 8}.
(b) A ∩ B = φ.
(c) C′ = {0, 1, 6, 7, 8, 9}.
(d) C′ ∩ D = {1, 6, 7}, so (C′ ∩ D) ∪ B = {1, 3, 5, 6, 7, 9}.
(e) (S ∩ C)′ = C′ = {0, 1, 6, 7, 8, 9}.
(f) A ∩ C = {2, 4}, so A ∩ C ∩ D′ = {2, 4}.
2.15 (a) A′ = {nitrogen, potassium, uranium, oxygen}.
(b) A ∪ C = {copper, sodium, zinc, oxygen}.
(c) A ∩ B′ = {copper, zinc} and
C′ = {copper, sodium, nitrogen, potassium, uranium, zinc};
so (A ∩ B′) ∪ C′ = {copper, sodium, nitrogen, potassium, uranium, zinc}.
14 Chapter 2 Probability
(d) B′ ∩ C′ = {copper, uranium, zinc}.
(e) A ∩ B ∩ C = φ.
(f) A′ ∪ B′ = {copper, nitrogen, potassium, uranium, oxygen, zinc} and
A′ ∩ C = {oxygen}; so, (A′ ∪ B′) ∩ (A′ ∩ C) = {oxygen}.
2.16 (a) M ∪ N = {x | 0 < x < 9}.
(b) M ∩ N = {x | 1 < x < 5}.
(c) M′ ∩ N′ = {x | 9 < x < 12}.
2.17 A Venn diagram is shown next.
A B
S
1 2 3 4
(a) From the above Venn diagram, (A ∩ B)′ contains the regions of 1, 2 and 4.
(b) (A ∪ B)′ contains region 1.
(c) A Venn diagram is shown next.
A
B
C
S
1
2
3
4
5
6
7
8
(A ∩ C) ∪ B contains the regions of 3, 4, 5, 7 and 8.
2.18 (a) Not mutually exclusive.
(b) Mutually exclusive.
(c) Not mutually exclusive.
(d) Mutually exclusive.
2.19 (a) The family will experience mechanical problems but will receive no ticket for
traffic violation and will not arrive at a campsite that has no vacancies.
(b) The family will receive a traffic ticket and arrive at a campsite that has no vacancies
but will not experience mechanical problems.
Solutions for Exercises in Chapter 2 15
(c) The family will experience mechanical problems and will arrive at a campsite that
has no vacancies.
(d) The family will receive a traffic ticket but will not arrive at a campsite that has
no vacancies.
(e) The family will not experience mechanical problems.
2.20 (a) 6;
(b) 2;
(c) 2, 5, 6;
(d) 4, 5, 6, 8.
2.21 With n1 = 6 sightseeing tours each available on n2 = 3 different days, the multiplication
rule gives n1n2 = (6)(3) = 18 ways for a person to arrange a tour.
2.22 With n1 = 8 blood types and n2 = 3 classifications of blood pressure, the multiplication
rule gives n1n2 = (8)(3) = 24 classifications.
2.23 Since the die can land in n1 = 6 ways and a letter can be selected in n2 = 26 ways, the
multiplication rule gives n1n2 = (6)(26) = 156 points in S.
2.24 Since a student may be classified according to n1 = 4 class standing and n2 = 2 gender
classifications, the multiplication rule gives n1n2 = (4)(2) = 8 possible classifications
for the students.
2.25 With n1 = 5 different shoe styles in n2 = 4 different colors, the multiplication rule
gives n1n2 = (5)(4) = 20 different pairs of shoes.
2.26 Using Theorem 2.8, we obtain the followings.
(a) There are
7
5
= 21 ways.
(b) There are
5
3
= 10 ways.
2.27 Using the generalized multiplication rule, there are n1×n2×n3×n4 = (4)(3)(2)(2) = 48
different house plans available.
2.28 With n1 = 5 different manufacturers, n2 = 3 different preparations, and n3 = 2
different strengths, the generalized multiplication rule yields n1n2n3 = (5)(3)(2) = 30
different ways to prescribe a drug for asthma.
2.29 With n1 = 3 race cars, n2 = 5 brands of gasoline, n3 = 7 test sites, and n4 = 2 drivers,
the generalized multiplication rule yields (3)(5)(7)(2) = 210 test runs.
2.30 With n1 = 2 choices for the first question, n2 = 2 choices for the second question,
and so forth, the generalized multiplication rule yields n1n2 · · ·n9 = 29 = 512 ways to
answer the test.
16 Chapter 2 Probability
2.31 (a) With n1 = 4 possible answers for the first question, n2 = 4 possible answers
for the second question, and so forth, the generalized multiplication rule yields
45 = 1024 ways to answer the test.
(b) With n1 = 3 wrong answers for the first question, n2 = 3 wrong answers for the
second question, and so forth, the generalized multiplication rule yields
n1n2n3n4n5 = (3)(3)(3)(3)(3) = 35 = 243
ways to answer the test and get all questions wrong.
2.32 (a) By Theorem 2.3, 7! = 5040.
(b) Since the first letter must be m, the remaining 6 letters can be arranged in 6! = 720
ways.
2.33 Since the first digit is a 5, there are n1 = 9 possibilities for the second digit and then
n2 = 8 possibilities for the third digit. Therefore, by the multiplication rule there are
n1n2 = (9)(8) = 72 registrations to be checked.
2.34 (a) By Theorem 2.3, there are 6! = 720 ways.
(b) A certain 3 persons can follow each other in a line of 6 people in a specified order is
4 ways or in (4)(3!) = 24 ways with regard to order. The other 3 persons can then
be placed in line in 3! = 6 ways. By Theorem 2.1, there are total (24)(6) = 144
ways to line up 6 people with a certain 3 following each other.
(c) Similar as in (b), the number of ways that a specified 2 persons can follow each
other in a line of 6 people is (5)(2!)(4!) = 240 ways. Therefore, there are 720 − 240 = 480 ways if a certain 2 persons refuse to follow each other.
2.35 The first house can be placed on any of the n1 = 9 lots, the second house on any of the
remaining n2 = 8 lots, and so forth. Therefore, there are 9! = 362, 880 ways to place
the 9 homes on the 9 lots.
2.36 (a) Any of the 6 nonzero digits can be chosen for the hundreds position, and of the
remaining 6 digits for the tens position, leaving 5 digits for the units position. So,
there are (6)(5)(5) = 150 three digit numbers.
(b) The units position can be filled using any of the 3 odd digits. Any of the remaining
5 nonzero digits can be chosen for the hundreds position, leaving a choice of 5
digits for the tens position. By Theorem 2.2, there are (3)(5)(5) = 75 three digit
odd numbers.
(c) If a 4, 5, or 6 is used in the hundreds position there remain 6 and 5 choices,
respectively, for the tens and units positions. This gives (3)(6)(5) = 90 three
digit numbers beginning with a 4, 5, or 6. If a 3 is used in the hundreds position,
then a 4, 5, or 6 must be used in the tens position leaving 5 choices for the units
position. In this case, there are (1)(3)(5) = 15 three digit number begin with
a 3. So, the total number of three digit numbers that are greater than 330 is
90 + 15 = 105.
Solutions for Exercises in Chapter 2 17
2.37 The first seat must be filled by any of 5 girls and the second seat by any of 4 boys.
Continuing in this manner, the total number of ways to seat the 5 girls and 4 boys is
(5)(4)(4)(3)(3)(2)(2)(1)(1) = 2880.
2.38 (a) 8! = 40320.
(b) There are 4! ways to seat 4 couples and then each member of a couple can be
interchanged resulting in 24(4!) = 384 ways.
(c) By Theorem 2.3, the members of each gender can be seated in 4! ways. Then
using Theorem 2.1, both men and women can be seated in (4!)(4!) = 576 ways.
2.39 (a) Any of the n1 = 8 finalists may come in first, and of the n2 = 7 remaining finalists
can then come in second, and so forth. By Theorem 2.3, there 8! = 40320 possible
orders in which 8 finalists may finish the spelling bee.
(b) The possible orders for the first three positions are 8P3 = 8!
5! = 336.
2.40 By Theorem 2.4, 8P5 = 8!
3! = 6720.
2.41 By Theorem 2.4, 6P4 = 6!
2! = 360.
2.42 By Theorem 2.4, 40P3 = 40!
37! = 59, 280.
2.43 By Theorem 2.5, there are 4! = 24 ways.
2.44 By Theorem 2.5, there are 7! = 5040 arrangements.
2.45 By Theorem 2.6, there are 8!
3!2! = 3360.
2.46 By Theorem 2.6, there are 9!
3!4!2! = 1260 ways.
2.47 By Theorem 2.7, there are
12
7,3,2
= 7920 ways.
2.48
9
1,4,4
+
9
2,4,3
+
9
1,3,5
+
9
2,3,4
+
9
2,2,5
= 4410.
2.49 By Theorem 2.8, there are
8
3
= 56 ways.
2.50 Assume February 29th as March 1st for the leap year. There are total 365 days in a
year. The number of ways that all these 60 students will have different birth dates (i.e,
arranging 60 from 365) is 365P60. This is a very large number.
2.51 (a) Sum of the probabilities exceeds 1.
(b) Sum of the probabilities is less than 1.
(c) A negative probability.
(d) Probability of both a heart and a black card is zero.
2.52 Assuming equal weights
18 Chapter 2 Probability
(a) P(A) = 5
18 ;
(b) P(C) = 1
3 ;
(c) P(A ∩ C) = 7
36 .
2.53 S = {$10, $25, $100} with weights 275/500 = 11/20, 150/500 = 3/10, and 75/500 =
3/20, respectively. The probability that the first envelope purchased contains less than
$100 is equal to 11/20 + 3/10 = 17/20.
2.54 (a) P(S ∩ D′) = 88/500 = 22/125.
(b) P(E ∩ D ∩ S′) = 31/500.
(c) P(S′ ∩ E′) = 171/500.
2.55 Consider the events
S: industry will locate in Shanghai,
B: industry will locate in Beijing.
(a) P(S ∩ B) = P(S) + P(B) − P(S ∪ B) = 0.7 + 0.4 − 0.8 = 0.3.
(b) P(S′ ∩ B′) = 1 − P(S ∪ B) = 1 − 0.8 = 0.2.
2.56 Consider the events
B: customer invests in tax-free bonds,
M: customer invests in mutual funds.
(a) P(B ∪M) = P(B) + P(M) − P(B ∩M) = 0.6 + 0.3 − 0.15 = 0.75.
(b) P(B′ ∩M′) = 1 − P(B ∪M) = 1 − 0.75 = 0.25.
2.57 (a) Since 5 of the 26 letters are vowels, we get a probability of 5/26.
(b) Since 9 of the 26 letters precede j, we get a probability of 9/26.
(c) Since 19 of the 26 letters follow g, we get a probability of 19/26.
2.58 (a) Let A = Defect in brake system; B = Defect in fuel system; P(A ∪ B) = P(A) +
P(B) − P(A ∩ B) = 0.25 + 0.17 − 0.15 = 0.27.
(b) P(No defect) = 1 − P(A ∪ B) = 1 − 0.27 = 0.73.
2.59 By Theorem 2.2, there are N = (26)(25)(24)(9)(8)(7)(6) = 47, 174, 400 possible ways
to code the items of which n = (5)(25)(24)(8)(7)(6)(4) = 4, 032, 000 begin with a vowel
and end with an even digit. Therefore, n
N = 10
117 .
2.60 (a) Of the (6)(6) = 36 elements in the sample space, only 5 elements (2,6), (3,5),
(4,4), (5,3), and (6,2) add to 8. Hence the probability of obtaining a total of 8 is
then 5/36.
(b) Ten of the 36 elements total at most 5. Hence the probability of obtaining a total
of at most is 10/36=5/18.
Solutions for Exercises in Chapter 2 19
2.61 Since there are 20 cards greater than 2 and less than 8, the probability of selecting two
of these in succession is
20
52
19
51
=
95
663
.
2.62 (a)
(1
1)(8
2)
(9
3)
= 1
3 .
(b)
(5
2)(3
1)
(9
3)
= 5
14 .
2.63 (a)
(4
3)(48
2 )
(52
5 )
= 94
54145 .
(b)
(13
4 )(13
1 )
(52
5 )
= 143
39984 .
2.64 Any four of a kind, say four 2’s and one 5 occur in
5
1
= 5 ways each with probability
(1/6)(1/6)(1/6)(1/6)(1/6) = (1/6)5. Since there are 6P2 = 30 ways to choose various
pairs of numbers to constitute four of one kind and one of the other (we use permutation
instead of combination is because that four 2’s and one 5, and four 5’s and one 2 are
two different ways), the probability is (5)(30)(1/6)5 = 25/1296.
2.65 (a) P(M ∪ H) = 88/100 = 22/25;
(b) P(M′ ∩ H′) = 12/100 = 3/25;
(c) P(H ∩M′) = 34/100 = 17/50.
2.66 (a) 9;
(b) 1/9.
2.67 (a) 0.32;
(b) 0.68;
(c) office or den.
2.68 (a) 1 − 0.42 = 0.58;
(b) 1 − 0.04 = 0.96.
2.69 P(A) = 0.2 and P(B) = 0.35
(a) P(A′) = 1 − 0.2 = 0.8;
(b) P(A′ ∩ B′) = 1 − P(A ∪ B) = 1 − 0.2 − 0.35 = 0.45;
(c) P(A ∪ B) = 0.2 + 0.35 = 0.55.
2.70 (a) 0.02 + 0.30 = 0.32 = 32%;
(b) 0.32 + 0.25 + 0.30 = 0.87 = 87%;
20 Chapter 2 Probability
(c) 0.05 + 0.06 + 0.02 = 0.13 = 13%;
(d) 1 − 0.05 − 0.32 = 0.63 = 63%.
2.71 (a) 0.12 + 0.19 = 0.31;
(b) 1 − 0.07 = 0.93;
(c) 0.12 + 0.19 = 0.31.
2.72 (a) 1 − 0.40 = 0.60.
(b) The probability that all six purchasing the electric oven or all six purchasing the
gas oven is 0.007 + 0.104 = 0.111. So the probability that at least one of each
type is purchased is 1 − 0.111 = 0.889.
2.73 (a) P(C) = 1 − P(A) − P(B) = 1 − 0.990 − 0.001 = 0.009;
(b) P(B′) = 1 − P(B) = 1 − 0.001 = 0.999;
(c) P(B) + P(C) = 0.01.
2.74 (a) ($4.50 − $4.00) × 50, 000 = $25, 000;
(b) Since the probability of underfilling is 0.001, we would expect 50, 000×0.001 = 50
boxes to be underfilled. So, instead of having ($4.50 − $4.00) × 50 = $25 profit
for those 50 boxes, there are a loss of $4.00 × 50 = $200 due to the cost. So, the
loss in profit expected due to underfilling is $25 + $200 = $250.
2.75 (a) 1 − 0.95 − 0.002 = 0.048;
(b) ($25.00 − $20.00) × 10, 000 = $50, 000;
(c) (0.05)(10, 000) × $5.00 + (0.05)(10, 000) × $20 = $12, 500.
2.76 P(A′∩B′) = 1−P(A∪B) = 1−(P(A)+P(B)−P(A∩B) = 1+P(A∩B)−P(A)−P(B).
2.77 (a) The probability that a convict who pushed dope, also committed armed robbery.
(b) The probability that a convict who committed armed robbery, did not push dope.
(c) The probability that a convict who did not push dope also did not commit armed
robbery.
2.78 P(S | A) = 10/18 = 5/9.
2.79 Consider the events:
M: a person is a male;
S: a person has a secondary education;
C: a person has a college degree.
(a) P(M | S) = 28/78 = 14/39;
(b) P(C′ | M′) = 95/112.
Solutions for Exercises in Chapter 2 21
2.80 Consider the events:
A: a person is experiencing hypertension,
B: a person is a heavy smoker,
C: a person is a nonsmoker.
(a) P(A | B) = 30/49;
(b) P(C | A′) = 48/93 = 16/31.
2.81 (a) P(M ∩ P ∩ H) = 10
68 = 5
34 ;
(b) P(H ∩M | P′) = P(H∩M∩P′)
P(P′) = 22−10
100−68 = 12
32 = 3
8 .
2.82 (a) (0.90)(0.08) = 0.072;
(b) (0.90)(0.92)(0.12) = 0.099.
2.83 (a) 0.018;
(b) 0.22 + 0.002 + 0.160 + 0.102 + 0.046 + 0.084 = 0.614;
(c) 0.102/0.614 = 0.166;
(d) 0.102+0.046
0.175+0.134 = 0.479.
2.84 Consider the events:
C: an oil change is needed,
F: an oil filter is needed.
(a) P(F | C) = P(F∩C)
P(C) = 0.14
0.25 = 0.56.
(b) P(C | F) = P(C∩F)
P(F) = 0.14
0.40 = 0.35.
2.85 Consider the events:
H: husband watches a certain show,
W: wife watches the same show.
(a) P(W ∩ H) = P(W)P(H | W) = (0.5)(0.7) = 0.35.
(b) P(W | H) = P(W∩H)
P(H) = 0.35
0.4 = 0.875.
(c) P(W ∪ H) = P(W) + P(H) − P(W ∩ H) = 0.5 + 0.4 − 0.35 = 0.55.
2.86 Consider the events:
H: the husband will vote on the bond referendum,
W: the wife will vote on the bond referendum.
Then P(H) = 0.21, P(W) = 0.28, and P(H ∩W) = 0.15.
(a) P(H ∪W) = P(H) + P(W) − P(H ∩W) = 0.21 + 0.28 − 0.15 = 0.34.
(b) P(W | H) = P(H∩W)
P(H) = 0.15
0.21 = 5
7 .
(c) P(H | W′) = P(H∩W′)
P(W′) = 0.06
0.72 = 1
12 .
22 Chapter 2 Probability
2.87 Consider the events:
A: the vehicle is a camper,
B: the vehicle has Canadian license plates.
(a) P(B | A) = P(A∩B)
P(A) = 0.09
0.28 = 9
28 .
(b) P(A | B) = P(A∩B)
P(B) = 0.09
0.12 = 3
4 .
(c) P(B′ ∪ A′) = 1 − P(A ∩ B) = 1 − 0.09 = 0.91.
2.88 Define
H: head of household is home,
C: a change is made in long distance carriers.
P(H ∩ C) = P(H)P(C | H) = (0.4)(0.3) = 0.12.
2.89 Consider the events:
A: the doctor makes a correct diagnosis,
B: the patient sues.
P(A′ ∩ B) = P(A′)P(B | A′) = (0.3)(0.9) = 0.27.
2.90 (a) 0.43;
(b) (0.53)(0.22) = 0.12;
(c) 1 − (0.47)(0.22) = 0.90.
2.91 Consider the events:
A: the house is open,
B: the correct key is selected.
P(A) = 0.4, P(A′) = 0.6, and P(B) =
(1
1)(7
2)
(8
3)
= 3
8 = 0.375.
So, P[A ∪ (A′ ∩ B)] = P(A) + P(A′)P(B) = 0.4 + (0.6)(0.375) = 0.625.
2.92 Consider the events:
F: failed the test,
P: passed the test.
(a) P(failed at least one tests) = 1 − P(P1P2P3P4) = 1 − (0.99)(0.97)(0.98)(0.99) =
1 − 0.93 = 0.07,
(b) P(failed 2 or 3) = P(P1)P(P4)(1 − P(P2P3)) = (0.99)(0.99)(1 − (0.97)(0.98)) =
0.0484.
(c) 100 × 0.07 = 7.
(d) 0.25.
2.93 Let A and B represent the availability of each fire engine.
(a) P(A′ ∩ B′) = P(A′)P(B′) = (0.04)(0.04) = 0.0016.
(b) P(A ∪ B) = 1 − P(A′ ∩ B′) = 1 − 0.0016 = 0.9984.
Solutions for Exercises in Chapter 2 23
2.94 P(T′ ∩ N′) = P(T′)P(N′) = (1 − P(T))(1 − P(N)) = (0.3)(0.1) = 0.03.
2.95 Consider the events:
A1: aspirin tablets are selected from the overnight case,
A2: aspirin tablets are selected from the tote bag,
L2: laxative tablets are selected from the tote bag,
T1: thyroid tablets are selected from the overnight case,
T2: thyroid tablets are selected from the tote bag.
(a) P(T1 ∩ T2) = P(T1)P(T2) = (3/5)(2/6) = 1/5.
(b) P(T′
1 ∩ T′
2) = P(T′
1)P(T′
2) = (2/5)(4/6) = 4/15.
(c) 1−P(A1 ∩A2)−P(T1 ∩T2) = 1−P(A1)P(A2)−P(T1)P(T2) = 1−(2/5)(3/6)−
(3/5)(2/6) = 3/5.
2.96 Consider the events:
X: a person has an X-ray,
C: a cavity is filled,
T: a tooth is extracted.
P(X ∩ C ∩ T) = P(X)P(C | X)P(T | X ∩ C) = (0.6)(0.3)(0.1) = 0.018.
2.97 (a) P(Q1 ∩Q2 ∩Q3 ∩Q4) = P(Q1)P(Q2 | Q1)P(Q3 | Q1 ∩Q2)P(Q4 | Q1 ∩Q2 ∩Q3) =
(15/20)(14/19)(13/18)(12/17) = 91/323.
(b) Let A be the event that 4 good quarts of milk are selected. Then
P(A) =
15
4
20
4
=
91
323
.
2.98 P = (0.95)[1 − (1 − 0.7)(1 − 0.8)](0.9) = 0.8037.
2.99 This is a parallel system of two series subsystems.
(a) P = 1 − [1 − (0.7)(0.7)][1 − (0.8)(0.8)(0.8)] = 0.75112.
(b) P = P(A′∩C∩D∩E)
Psystem works = (0.3)(0.8)(0.8)(0.8)
0.75112 = 0.2045.
2.100 Define S: the system works.
P(A′ | S′) = P(A′∩S′)
P(S′) = P(A′)(1−P(C∩D∩E))
1−P(S) = (0.3)[1−(0.8)(0.8)(0.8)]
1−0.75112 = 0.588.
2.101 Consider the events:
C: an adult selected has cancer,
D: the adult is diagnosed as having cancer.
P(C) = 0.05, P(D | C) = 0.78, P(C′) = 0.95 and P(D | C′) = 0.06. So, P(D) =
P(C ∩ D) + P(C′ ∩ D) = (0.05)(0.78) + (0.95)(0.06) = 0.096.
24 Chapter 2 Probability
2.102 Let S1, S2, S3, and S4 represent the events that a person is speeding as he passes through
the respective locations and let R represent the event that the radar traps is operating
resulting in a speeding ticket. Then the probability that he receives a speeding ticket:
P(R) =
P4
i=1
P(R | Si)P(Si) = (0.4)(0.2) + (0.3)(0.1) + (0.2)(0.5) + (0.3)(0.2) = 0.27.
2.103 P(C | D) = P(C∩D)
P(D) = 0.039
0.096 = 0.40625.
2.104 P(S2 | R) = P(R∩ S2)
P(R) = 0.03
0.27 = 1/9.
2.105 Consider the events:
A: no expiration date,
B1: John is the inspector, P(B1) = 0.20 and P(A | B1) = 0.005,
B2: Tom is the inspector, P(B2) = 0.60 and P(A | B2) = 0.010,
B3: Jeff is the inspector, P(B3) = 0.15 and P(A | B3) = 0.011,
B4: Pat is the inspector, P(B4) = 0.05 and P(A | B4) = 0.005,
P(B1 | A) = (0.005)(0.20)
(0.005)(0.20)+(0.010)(0.60)+(0.011)(0.15)+(0.005)(0.05) = 0.1124.
2.106 Consider the events
E: a malfunction by other human errors,
A: station A, B: station B, and C: station C.
P(C | E) = P(E | C)P(C)
P(E | A)P(A)+P(E | B)P(B)+P(E | C)P(C) = (5/10)(10/43)
(7/18)(18/43)+(7/15)(15/43)+(5/10)(10/43) =
0.1163
0.4419 = 0.2632.
2.107 (a) P(A ∩ B ∩ C) = P(C | A ∩ B)P(B | A)P(A) = (0.20)(0.75)(0.3) = 0.045.
(b) P(B′ ∩ C) = P(A ∩ B′ ∩ C) + P(A′ ∩ B′ ∩ C) = P(C | A ∩ B′)P(B′ | A)P(A) +
P(C | A′∩B′)P(B′ | A′)P(A′) = (0.80)(1−0.75)(0.3)+(0.90)(1−0.20)(1−0.3) =
0.564.
(c) Use similar argument as in (a) and (b), P(C) = P(A∩B ∩C)+P(A∩B′ ∩ C)+
P(A′ ∩ B ∩ C) + P(A′ ∩ B′ ∩ C) = 0.045 + 0.060 + 0.021 + 0.504 = 0.630.
(d) P(A | B′ ∩ C) = P(A ∩ B′ ∩ C)/P(B′ ∩ C) = (0.06)(0.564) = 0.1064.
2.108 Consider the events:
A: a customer purchases latex paint,
A′: a customer purchases semigloss paint,
B: a customer purchases rollers.
P(A | B) = P(B | A)P(A)
P(B | A)P(A)+P(B | A′)P(A′) = (0.60)(0.75)
(0.60)(0.75)+(0.25)(0.30) = 0.857.
2.109 Consider the events:
G: guilty of committing a crime,
I: innocent of the crime,
i: judged innocent of the crime,
g: judged guilty of the crime.
P(I | g) = P(g | I)P(I)
P(g | G)P(G)+P(g | I)P(I) = (0.01)(0.95)
(0.05)(0.90)+(0.01)(0.95) = 0.1743.
Solutions for Exercises in Chapter 2 25
2.110 Let Ai be the event that the ith patient is allergic to some type of week.
(a) P(A1 ∩ A2 ∩ A3 ∩ A′
4) + P(A1 ∩ A2 ∩ A′
3 ∩ A4) + P(A1 ∩ A′
2 ∩ A3 ∩ A4) +
P(A′
1 ∩ A2 ∩ A3 ∩ A4) = P(A1)P(A2)P(A3)P(A′
4) + P(A1)P(A2)P(A′
3)P(A4) +
P(A1)P(A′
2)P(A3)P(A4) + P(A′
1)P(A2)P(A3)P(A4) = (4)(1/2)4 = 1/4.
(b) P(A′
1 ∩ A′
2 ∩ A′
3 ∩ A′
4) = P(A′
1)P(A′
2)P(A′
3)P(A′
4) = (1/2)4 = 1/16.
2.111 No solution necessary.
2.112 (a) 0.28 + 0.10 + 0.17 = 0.55.
(b) 1 − 0.17 = 0.83.
(c) 0.10 + 0.17 = 0.27.
2.113 P =
(13
4 )(13
6 )(13
1 )(13
2 )
(52
13)
.
2.114 (a) P(M1 ∩M2 ∩M3 ∩M4) = (0.1)4 = 0.0001, where Mi represents that ith person
make a mistake.
(b) P(J ∩ C ∩ R′ ∩W′) = (0.1)(0.1)(0.9)(0.9) = 0.0081.
2.115 Let R, S, and L represent the events that a client is assigned a room at the Ramada
Inn, Sheraton, and Lakeview Motor Lodge, respectively, and let F represents the event
that the plumbing is faulty.
(a) P(F) = P(F | R)P(R) + P(F | S)P(S) + P(F | L)P(L) = (0.05)(0.2) +
(0.04)(0.4) + (0.08)(0.3) = 0.054.
(b) P(L | F) = (0.08)(0.3)
0.054 = 4
9 .
2.116 (a) There are
9
3
= 84 possible committees.
(b) There are
4
1
5
2
= 40 possible committees.
(c) There are
3
1
1
1
5
1
= 15 possible committees.
2.117 Denote by R the event that a patient survives. Then P(R) = 0.8.
(a) P(R1 ∩ R2 ∩ R′
3) + P(R1 ∩ R′
2 ∩ R3)P(R′
1 ∩ R2 ∩ R3) = P(R1)P(R2)P(R′
3) +
P(R1)P(R′
2)P(R3) + P(R′
1)P(R2)P(R3) = (3)(0.8)(0.8)(0.2) = 0.384.
(b) P(R1 ∩ R2 ∩ R3) = P(R1)P(R2)P(R3) = (0.8)3 = 0.512.
2.118 Consider events
M: an inmate is a male,
N: an inmate is under 25 years of age.
P(M′ ∩ N′) = P(M′) + P(N′) − P(M′ ∪ N′) = 2/5 + 1/3 − 5/8 = 13/120.
2.119 There are
4
3
5
3
6
3
= 800 possible selections.
26 Chapter 2 Probability
2.120 Consider the events:
Bi: a black ball is drawn on the ith drawl,
Gi: a green ball is drawn on the ith drawl.
(a) P(B1 ∩B2 ∩B3)+P(G1 ∩G2 ∩G3) = (6/10)(6/10)(6/10)+(4/10)(4/10)(4/10) =
7/25.
(b) The probability that each color is represented is 1 − 7/25 = 18/25.
2.121 The total number of ways to receive 2 or 3 defective sets among 5 that are purchased
is
3
2
9
3
+
3
3
9
2
= 288.
2.122 A Venn diagram is shown next.
A
B
C
S
1
2
3
4
5
6
7
8
(a) (A ∩ B)′: 1, 2, 3, 6, 7, 8.
(b) (A ∪ B)′: 1, 6.
(c) (A ∩ C) ∪ B: 3, 4, 5, 7, 8.
2.123 Consider the events:
O: overrun,
A: consulting firm A,
B: consulting firm B,
C: consulting firm C.
(a) P(C | O) = P(O | C)P(C)
P(O | A)P(A)+P(O | B)P(B)+P(O | C)P(C) = (0.15)(0.25)
(0.05)(0.40)+(0.03)(0.35)+(0.15)(0.25) =
0.0375
0.0680 = 0.5515.
(b) P(A | O) = (0.05)(0.40)
0.0680 = 0.2941.
2.124 (a) 36;
(b) 12;
(c) order is not important.
2.125 (a) 1
(36
2 )
= 0.0016;
(b)
(12
1 )(24
1 )
(36
2 )
= 288
630 = 0.4571.
Solutions for Exercises in Chapter 2 27
2.126 Consider the events:
C: a woman over 60 has the cancer,
P: the test gives a positive result.
So, P(C) = 0.07, P(P′ | C) = 0.1 and P(P | C′) = 0.05.
P(C | P′) = P(P′ | C)P(C)
P(P′ | C)P(C)+P(P′ | C′)P(C′) = (0.1)(0.07)
(0.1)(0.07)+(1−0.05)(1−0.07) = 0.007
0.8905 = 0.00786.
2.127 Consider the events:
A: two nondefective components are selected,
N: a lot does not contain defective components, P(N) = 0.6, P(A | N) = 1,
O: a lot contains one defective component, P(O) = 0.3, P(A | O) =
(19
2 )
(20
2 )
= 9
10 ,
T: a lot contains two defective components,P(T) = 0.1, P(A | T) =
(18
2 )
(20
2 )
= 153
190 .
(a) P(N | A) = P(A | N)P(N)
P(A | N)P(N)+P(A | O)P(O)+P(A | T)P(T) = (1)(0.6)
(1)(0.6)+(9/10)(0.3)+(153/190)(0.1)
= 0.6
0.9505 = 0.6312;
(b) P(O | A) = (9/10)(0.3)
0.9505 = 0.2841;
(c) P(T | A) = 1 − 0.6312 − 0.2841 = 0.0847.
2.128 Consider events:
D: a person has the rare disease, P(D) = 1/500,
P: the test shows a positive result, P(P | D) = 0.95 and P(P | D′) = 0.01.
P(D | P) = P(P | D)P(D)
P(P | D)P(D)+P(P | D′)P(D′) = (0.95)(1/500)
(0.95)(1/500)+(0.01)(1−1/500) = 0.1599.
2.129 Consider the events:
1: engineer 1, P(1) = 0.7, and 2: engineer 2, P(2) = 0.3,
E: an error has occurred in estimating cost, P(E | 1) = 0.02 and P(E | 2) = 0.04.
P(1 | E) = P(E | 1)P(1)
P(E | 1)P(1)+P(E | 2)P(2) = (0.02)(0.7)
(0.02)(0.7)+(0.04)(0.3) = 0.5385, and
P(2 | E) = 1 − 0.5385 = 0.4615. So, more likely engineer 1 did the job.
2.130 Consider the events: D: an item is defective
(a) P(D1D2D3) = P(D1)P(D2)P(D3) = (0.2)3 = 0.008.
(b) P(three out of four are defectives) =
4
3
(0.2)3(1 − 0.2) = 0.0256.
2.131 Let A be the event that an injured worker is admitted to the hospital and N be the event
that an injured worker is back to work the next day. P(A) = 0.10, P(N) = 0.15 and
P(A∩N) = 0.02. So, P(A∪N) = P(A)+P(N)−P(A∩N) = 0.1+0.15−0.02 = 0.23.
2.132 Consider the events:
T: an operator is trained, P(T) = 0.5,
M an operator meets quota, P(M | T) = 0.9 and P(M | T′) = 0.65.
P(T | M) = P(M | T)P(T)
P(M | T)P(T)+P(M | T′)P(T′) = (0.9)(0.5)
(0.9)(0.5)+(0.65)(0.5) = 0.5807.
28 Chapter 2 Probability
2.133 Consider the events:
A: purchased from vendor A,
D: a customer is dissatisfied.
Then P(A) = 0.2, P(A | D) = 0.5, and P(D) = 0.1.
So, P(D | A) = P(A | D)P(D)
P(A) = (0.5)(0.1)
0.2 = 0.25.
2.134 (a) P(Union member | New company (same field)) = 13
13+10 = 13
23 = 0.5652.
(b) P(Unemployed | Union member) = 2
40+13+4+2 = 2
59 = 0.034.
2.135 Consider the events:
C: the queen is a carrier, P(C) = 0.5,
D: a prince has the disease, P(D | C) = 0.5.
P(C | D′
1D′
2D′
3) = P(D
′
1D
′
2D
′
3 | C)P(C)
P(D
′
1D
′
2D
′
3 | C)P(C)+P(D
′
1D
′
2D
′
3 | C′)P(C′)
= (0.5)3(0.5)
(0.5)3(0.5)+1(0.5) = 1
9 .
2.136 Using the solutions to Exercise 2.50, we know that there are total 365P60 ways that no
two students have the same birth date. Since the total number of ways of the birth
dates that 60 students can have is 36560, the probability that at least two students
will have the same birth date in a class of 60 is P = 1 − 365P60
36560 . To compute this
number, regular calculator may not be able to handle it. Using approximation (such
as Stirling’s approximation formula), we obtain P = 0.9941, which is quite high.
Chapter 3
Random Variables and Probability
Distributions
3.1 Discrete; continuous; continuous; discrete; discrete; continuous.
3.2 A table of sample space and assigned values of the random variable is shown next.
Sample Space x
NNN
NNB
NBN
BNN
NBB
BNB
BBN
BBB
0
1
1
1
2
2
2
3
3.3 A table of sample space and assigned values of the random variable is shown next.
Sample Space w
HHH
HHT
HTH
THH
HTT
THT
TTH
TTT
3
1
1
1
−1
−1
−1
−3
3.4 S = {HHH, THHH,HTHHH, TTHHH, TTTHHH,HTTHHH, THTHHH,
HHTHHH, . . . }; The sample space is discrete containing as many elements as there
are positive integers.
29
30 Chapter 3 Random Variables and Probability Distributions
3.5 (a) c = 1/30 since 1 =
P3
x=0
c(x2 + 4) = 30c.
(b) c = 1/10 since
1 =
X2
x=0
c
2
x
3
3 − x
= c
2
0
3
3
+
2
1
3
2
+
2
2
3
1
= 10c.
3.6 (a) P(X > 200) =
R
∞
200
20000
(x+100)3 dx = − 10000
(x+100)2
∞
200
= 1
9 .
(b) P(80 < X < 200) =
R 120
80
20000
(x+100)3 dx = − 10000
(x+100)2
120
80
= 1000
9801 = 0.1020.
3.7 (a) P(X < 1.2) =
R 1
0 x dx +
R 1.2
1 (2 − x) dx = x2
2
1
0
+
2x − x2
2
1.2
1
= 0.68.
(b) P(0.5 < X < 1) =
R 1
0.5 x dx = x2
2
1
0.5
= 0.375.
3.8 Referring to the sample space in Exercise 3.3 and making use of the fact that P(H) =
2/3 and P(T) = 1/3, we have
P(W = −3) = P(TTT) = (1/3)3 = 1/27;
P(W = −1) = P(HTT) + P(THT) + P(TTH) = 3(2/3)(1/3)2 = 2/9;
P(W = 1) = P(HHT) + P(HTH) + P(THH) = 3(2/3)2(1/3) = 2/9;
P(W = 3) = P(HHH) = (2/3)3 = 8/27;
The probability distribution for W is then
w −3 −1 1 3
P(W = w) 1/27 2/9 2/9 8/27
3.9 (a) P(0 < X < 1) =
R 1
0
2(x+2)
5 dx = (x+2)2
5
1
0
= 1.
(b) P(1/4 < X < 1/2) =
R 1/2
1/4
2(x+2)
5 dx = (x+2)2
5
1/2
1/4
= 19/80.
3.10 The die can land in 6 different ways each with probability 1/6. Therefore, f(x) = 1
6 ,
for x = 1, 2, . . . , 6.
3.11 We can select x defective sets from 2, and 3 − x good sets from 5 in
2
x
5
3−x
ways. A
random selection of 3 from 7 sets can be made in
7
3
ways. Therefore,
f(x) =
2
x
5
3−x
7
3
, x = 0, 1, 2.
In tabular form
x 0 1 2
f(x) 2/7 4/7 1/7
Solutions for Exercises in Chapter 3 31
The following is a probability histogram:
1 2 3
x
f(x)
1/7
2/7
3/7
4/7
3.12 (a) P(T = 5) = F(5) − F(4) = 3/4 − 1/2 = 1/4.
(b) P(T > 3) = 1 − F(3) = 1 − 1/2 = 1/2.
(c) P(1.4 < T < 6) = F(6) − F(1.4) = 3/4 − 1/4 = 1/2.
3.13 The c.d.f. of X is
F(x) =
0, for x < 0,
0.41, for 0 ≤ x < 1,
0.78, for 1 ≤ x < 2,
0.94, for 2 ≤ x < 3,
0.99, for 3 ≤ x < 4,
1, for x ≥ 4.
3.14 (a) P(X < 0.2) = F(0.2) = 1 − e−1.6 = 0.7981;
(b) f(x) = F′(x) = 8e−8x. Therefore, P(X < 0.2) = 8
R 0.2
0 e−8x dx = −e−8x|0.2
0 =
0.7981.
3.15 The c.d.f. of X is
F(x) =
0, for x < 0,
2/7, for 0 ≤ x < 1,
6/7, for 1 ≤ x < 2,
1, for x ≥ 2.
(a) P(X = 1) = P(X ≤ 1) − P(X ≤ 0) = 6/7 − 2/7 = 4/7;
(b) P(0 < X ≤ 2) = P(X ≤ 2) − P(X ≤ 0) = 1 − 2/7 = 5/7.
32 Chapter 3 Random Variables and Probability Distributions
3.16 A graph of the c.d.f. is shown next.
F(x)
x
1/7
2/7
3/7
4/7
5/7
6/7
1
0 1 2
3.17 (a) Area =
R 3
1 (1/2) dx = x
2
3
1 = 1.
(b) P(2 < X < 2.5)
R 2.5
2 (1/2) dx = x
2
2.5
2 = 1
4 .
(c) P(X ≤ 1.6) =
R 1.6
1 (1/2) dx = x
2
1
.
6
1 = 0.3.
3.18 (a) P(X < 4) =
R 4
2
2(1+x)
27 dx = (1+x)2
27
4
2
= 16/27.
(b) P(3 ≤ X < 4) =
R 4
3
2(1+x)
27 dx = (1+x)2
27
4
3
= 1/3.
3.19 F(x) =
R x
1 (1/2) dt = x−1
2 ,
P(2 < X < 2.5) = F(2.5) − F(2) = 1.5
2 − 1
2 = 1
4 .
3.20 F(x) = 2
27
R x
2 (1 + t) dt = 2
27
t + t2
2
x
2
= (x+4)(x−2)
27 ,
P(3 ≤ X < 4) = F(4) − F(3) = (8)(2)
27 − (7)(1)
27 = 1
3 .
3.21 (a) 1 = k
R 1
0 √x dx = 2k
3 x3/2
1
0 = 2k
3 . Therefore, k = 3
2 .
(b) F(x) = 3
2
R x
0
√t dt = t3/2
x
0 = x3/2.
P(0.3 < X < 0.6) = F(0.6) − F(0.3) = (0.6)3/2 − (0.3)3/2 = 0.3004.
3.22 Denote by X the number of spades int he three draws. Let S and N stand for a spade
and not a spade, respectively. Then
P(X = 0) = P(NNN) = (39/52)(38/51)(37/50) = 703/1700,
P(X = 1) = P(SNN) + P(NSN) + P(NNS) = 3(13/52)(39/51)(38/50) = 741/1700,
P(X = 3) = P(SSS) = (13/52)(12/51)(11/50) = 11/850, and
P(X = 2) = 1 − 703/1700 − 741/1700 − 11/850 = 117/850.
The probability mass function for X is then
x 0 1 2 3
f(x) 703/1700 741/1700 117/850 11/850
Solutions for Exercises in Chapter 3 33
3.23 The c.d.f. of X is
F(x) =
0, for w < −3,
1/27, for − 3 ≤ w < −1,
7/27, for − 1 ≤ w < 1,
19/27, for 1 ≤ w < 3,
1, for w ≥ 3,
(a) P(W > 0 = 1 − P(W ≤ 0) = 1 − 7/27 = 20/27.
(b) P(−1 ≤ W < 3) = F(2) − F(−3) = 19/27 − 1/27 = 2/3.
3.24 There are
10
4
ways of selecting any 4 CDs from 10. We can select x jazz CDs from 5
and 4 − x from the remaining CDs in
5
x
5
4−x
ways. Hence
f(x) =
5
x
5
4−x
10
4
, x = 0, 1, 2, 3, 4.
3.25 Let T be the total value of the three coins. Let D and N stand for a dime and nickel,
respectively. Since we are selecting without replacement, the sample space containing
elements for which t = 20, 25, and 30 cents corresponding to the selecting of 2 nickels
and 1 dime, 1 nickel and 2 dimes, and 3 dimes. Therefore, P(T = 20) =
(2
2)(4
1)
(6
3)
= 1
5 ,
P(T = 25) =
(2
1)(4
2)
(6
3)
= 3
5 ,
P(T = 30) =
(4
3)
(6
3)
= 1
5 ,
and the probability distribution in tabular form is
t 20 25 30
P(T = t) 1/5 3/5 1/5
As a probability histogram
20 25 30
x
f(x)
1/5
2/5
3/5
34 Chapter 3 Random Variables and Probability Distributions
3.26 Denote by X the number of green balls in the three draws. Let G and B stand for the
colors of green and black, respectively.
Simple Event x P(X = x)
BBB
GBB
BGB
BBG
BGG
GBG
GGB
GGG
0
1
1
1
2
2
2
3
(2/3)3 = 8/27
(1/3)(2/3)2 = 4/27
(1/3)(2/3)2 = 4/27
(1/3)(2/3)2 = 4/27
(1/3)2(2/3) = 2/27
(1/3)2(2/3) = 2/27
(1/3)2(2/3) = 2/27
(1/3)3 = 1/27
The probability mass function for X is then
x 0 1 2 3
P(X = x) 8/27 4/9 2/9 1/27
3.27 (a) For x ≥ 0, F(x) =
R x
0
1
2000 exp(−t/2000) dt = − exp(−t/2000)|x
0
= 1 − exp(−x/2000). So
F(x) =
(
0, x < 0,
1 − exp(−x/2000), x ≥ 0.
(b) P(X > 1000) = 1 − F(1000) = 1 − [1 − exp(−1000/2000)] = 0.6065.
(c) P(X < 2000) = F(2000) = 1 − exp(−2000/2000) = 0.6321.
3.28 (a) f(x) ≥ 0 and
R 26.25
23.75
2
5 dx = 2
5 t
26.25
23.75 = 2.5
2.5 = 1.
(b) P(X < 24) =
R 24
23.75
2
5 dx = 2
5 (24 − 23.75) = 0.1.
(c) P(X > 26) =
R 26.25
26
2
5 dx = 2
5 (26.25 − 26) = 0.1. It is not extremely rare.
3.29 (a) f(x) ≥ 0 and
R
∞
1 3x−4 dx = −3 x−3
3
∞
1
= 1. So, this is a density function.
(b) For x ≥ 1, F(x) =
R x
1 3t−4 dt = 1 − x−3. So,
F(x) =
(
0, x < 1,
1 − x−3, x ≥ 1.
(c) P(X > 4) = 1 − F(4) = 4−3 = 0.0156.
3.30 (a) 1 = k
R 1
−1(3 − x2) dx = k
3x − x3
3
1
−1
= 16
3 k. So, k = 3
16 .
Solutions for Exercises in Chapter 3 35
(b) For −1 ≤ x < 1, F(x) = 3
16
R x
−1(3 − t2) dt =
3t − 1
3 t3
x
−1 = 1
2 + 9
16x − x3
16 .
So, P
X < 1
2
= 1
2 −
9
16
1
2
− 1
16
1
2
3
= 99
128 .
(c) P(|X| < 0.8) = P(X < −0.8) + P(X > 0.8) = F(−0.8) + 1 − F(0.8)
= 1 +
1
2 − 9
160.8 + 1
160.83
−
1
2 + 9
160.8 − 1
160.83
= 0.164.
3.31 (a) For y ≥ 0, F(y) = 1
4
R y
0 e−t/4 dy = 1 − ey/4. So, P(Y > 6) = e−6/4 = 0.2231. This
probability certainly cannot be considered as “unlikely.”
(b) P(Y ≤ 1) = 1 − e−1/4 = 0.2212, which is not so small either.
3.32 (a) f(y) ≥ 0 and
R 1
0 5(1 − y)4 dy = − (1 − y)5|1
0 = 1. So, this is a density function.
(b) P(Y < 0.1) = − (1 − y)5|0.1
0 = 1 − (1 − 0.1)5 = 0.4095.
(c) P(Y > 0.5) = (1 − 0.5)5 = 0.03125.
3.33 (a) Using integral by parts and setting 1 = k
R 1
0 y4(1 − y)3 dy, we obtain k = 280.
(b) For 0 ≤ y < 1, F(y) = 56y5(1 − Y )3 + 28y6(1 − y)2 + 8y7(1 − y) + y8. So,
P(Y ≤ 0.5) = 0.3633.
(c) Using the cdf in (b), P(Y > 0.8) = 0.0563.
3.34 (a) The event Y = y means that among 5 selected, exactly y tubes meet the specification
(M) and 5 − y (M′) does not. The probability for one combination of
such a situation is (0.99)y(1 − 0.99)5−y if we assume independence among the
tubes. Since there are 5!
y!(5−y)! permutations of getting y Ms and 5 − y M′s, the
probability of this event (Y = y) would be what it is specified in the problem.
(b) Three out of 5 is outside of specification means that Y = 2. P(Y = 2) = 9.8×10−6
which is extremely small. So, the conjecture is false.
3.35 (a) P(X > 8) = 1 − P(X ≤ 8) =
P8
x=0
e−6 6x
x! = e−6
60
0! + 61
1! + · · · + 68
8!
= 0.1528.
(b) P(X = 2) = e−6 62
2! = 0.0446.
3.36 For 0 < x < 1, F(x) = 2
R x
0 (1 − t) dt = − (1 − t)2|x
0 = 1 − (1 − x)2.
(a) P(X ≤ 1/3) = 1 − (1 − 1/3)2 = 5/9.
(b) P(X > 0.5) = (1 − 1/2)2 = 1/4.
(c) P(X < 0.75 | X ≥ 0.5) = P(0.5≤X<0.75)
P(X≥0.5) = (1−0.5)2−(1−0.75)2
(1−0.5)2 = 3
4 .
3.37 (a)
P3
x=0
P3
y=0
f(x, y) = c
P3
x=0
P3
y=0
xy = 36c = 1. Hence c = 1/36.
(b)
P
x
P
y
f(x, y) = c
P
x
P
y |x − y| = 15c = 1. Hence c = 1/15.
3.38 The joint probability distribution of (X, Y ) is
36 Chapter 3 Random Variables and Probability Distributions
x
f(x, y) 0 1 2 3
0 0 1/30 2/30 3/30
y 1 1/30 2/30 3/30 4/30
2 2/30 3/30 4/30 5/30
(a) P(X ≤ 2, Y = 1) = f(0, 1) + f(1, 1) + f(2, 1) = 1/30 + 2/30 + 3/30 = 1/5.
(b) P(X > 2, Y ≤ 1) = f(3, 0) + f(3, 1) = 3/30 + 4/30 = 7/30.
(c) P(X > Y ) = f(1, 0) + f(2, 0) + f(3, 0) + f(2, 1) + f(3, 1) + f(3, 2)
= 1/30 + 2/30 + 3/30 + 3/30 + 4/30 + 5/30 = 3/5.
(d) P(X + Y = 4) = f(2, 2) + f(3, 1) = 4/30 + 4/30 = 4/15.
3.39 (a) We can select x oranges from 3, y apples from 2, and 4 − x − y bananas from 3
in
3
x
2
y
3
4−x−y
ways. A random selection of 4 pieces of fruit can be made in
8
4
ways. Therefore,
f(x, y) =
3
x
2
y
3
4−x−y
8
4
, x = 0, 1, 2, 3; y = 0, 1, 2; 1 ≤ x + y ≤ 4.
(b) P[(X, Y ) ∈ A] = P(X + Y ≤ 2) = f(1, 0) + f(2, 0) + f(0, 1) + f(1, 1) + f(0, 2)
= 3/70 + 9/70 + 2/70 + 18/70 + 3/70 = 1/2.
3.40 (a) g(x) = 2
3
R 1
0 (x + 2y) dy = 2
3 (x + 1), for 0 ≤ x ≤ 1.
(b) h(y) = 2
3
R 1
0 (x + 2y) dy = 1
3 (1 + 4y), for 0 ≤ y ≤ 1.
(c) P(X < 1/2) = 2
3
R 1/2
0 (x + 1) dx = 5
12 .
3.41 (a) P(X + Y ≤ 1/2) =
R 1/2
0
R 1/2−y
0 24xy dx dy = 12
R 1/2
0
1
2 − y
2
y dy = 1
16 .
(b) g(x) =
R 1−x
0 24xy dy = 12x(1 − x)2, for 0 ≤ x < 1.
(c) f(y|x) = 24xy
12x(1−x)2 = 2y
(1−x)2 , for 0 ≤ y ≤ 1 − x.
Therefore, P(Y < 1/8 | X = 3/4) = 32
R 1/8
0 y dy = 1/4.
3.42 Since h(y) = e−y
R
∞
0 e−x dx = e−y, for y > 0, then f(x|y) = f(x, y)/h(y) = e−x, for
x > 0. So, P(0 < X < 1 | Y = 2) =
R 1
0 e−x dx = 0.6321.
3.43 (a) P(0 ≤ X ≤ 1/2, 1/4 ≤ Y ≤ 1/2) =
R 1/2
0
R 1/2
1/4 4xy dy dx = 3/8
R 1/2
0 x dx = 3/64.
(b) P(X < Y ) =
R 1
0
R y
0 4xy dx dy = 2
R 1
0 y3 dy = 1/2.
3.44 (a) 1 = k
R 50
30
R 50
30 (x2 + y2) dx dy = k(50 − 30)
R 50
30 x2 dx +
R 50
30 y2 dy
= 392k
3 · 104.
So, k = 3
392 · 10−4.
Solutions for Exercises in Chapter 3 37
(b) P(30 ≤ X ≤ 40, 40 ≤ Y ≤ 50) = 3
392 · 10−4
R 40
30
R 50
40 (x2 + y2) dy dx
= 3
392 · 10−3(
R 40
30 x2 dx +
R 50
40 y2 dy) = 3
392 · 10−3
403−303
3 + 503−403
3
= 49
196 .
(c) P(30 ≤ X ≤ 40, 30 ≤ Y ≤ 40) = 3
392 · 10−4
R 40
30
R 40
30 (x2 + y2) dx dy
= 2 3
392 · 10−4(40 − 30)
R 40
30 x2 dx = 3
196 · 10−3 403−303
3 = 37
196 .
3.45 P(X + Y > 1/2) = 1 − P(X + Y < 1/2) = 1 −
R 1/4
0
R 1/2−x
x
1
y dy dx
= 1 −
R 1/4
0
ln
1
2 − x
− ln x
dx = 1 +
1
2 − x
ln
1
2 − x
− x ln x
1/4
0
= 1 + 1
4 ln
1
4
= 0.6534.
3.46 (a) From the column totals of Exercise 3.38, we have
x 0 1 2 3
g(x) 1/10 1/5 3/10 2/5
(b) From the row totals of Exercise 3.38, we have
y 0 1 2
h(y) 1/5 1/3 7/15
3.47 (a) g(x) = 2
R 1
x dy = 2(1 − x) for 0 < x < 1;
h(y) = 2
R y
0 dx = 2y, for 0 < y < 1.
Since f(x, y) 6= g(x)h(y), X and Y are not independent.
(b) f(x|y) = f(x, y)/h(y) = 1/y, for 0 < x < y.
Therefore, P(1/4 < X < 1/2 | Y = 3/4) = 4
3
R 1/2
1/4 dx = 1
3 .
3.48 (a) g(2) =
P2
y=0
f(2, y) = f(2, 0) + f(2, 1) + f(2, 2) = 9/70 + 18/70 + 3/70 = 3/7. So,
f(y|2) = f(2, y)/g(2) = (7/3)f(2, y).
f(0|2) = (7/3)f(2, 0) = (7/3)(9/70) = 3/10, f(1|2) = 3/5 and f(2|2) = 1/10. In
tabular form,
y 0 1 2
f(y|2) 3/10 3/5 1/10
(b) P(Y = 0 | X = 2) = f(0|2) = 3/10.
3.49 (a)
x 1 2 3
g(x) 0.10 0.35 0.55
(b)
y 1 2 3
h(y) 0.20 0.50 0.30
(c) P(Y = 3 | X = 2) = 0.2
0.05+0.10+0.20 = 0.5714.
38 Chapter 3 Random Variables and Probability Distributions
3.50
x
f(x, y) 2 4 h(y)
1 0.10 0.15 0.25
y 3 0.20 0.30 0.50
5 0.10 0.15 0.25
g(x) 0.40 0.60
(a)
x 2 4
g(x) 0.40 0.60
(b)
y 1 3 5
h(y) 0.25 0.50 0.25
3.51 (a) Let X be the number of 4’s and Y be the number of 5’s. The sample space
consists of 36 elements each with probability 1/36 of the form (m, n) where
m is the outcome of the first roll of the die and n is the value obtained on
the second roll. The joint probability distribution f(x, y) is defined for x =
0, 1, 2 and y = 0, 1, 2 with 0 ≤ x + y ≤ 2. To find f(0, 1), for example,
consider the event A of obtaining zero 4’s and one 5 in the 2 rolls. Then
A = {(1, 5), (2, 5), (3, 5), (6, 5), (5, 1), (5, 2), (5, 3), (5, 6)}, so f(0, 1) = 8/36 = 2/9.
In a like manner we find f(0, 0) = 16/36 = 4/9, f(0, 2) = 1/36, f(1, 0) = 2/9,
f(2, 0) = 1/36, and f(1, 1) = 1/18.
(b) P[(X, Y ) ∈ A] = P(2X + Y < 3) = f(0, 0) + f(0, 1) + f(0, 2) + f(1, 0) =
4/9 + 1/9 + 1/36 + 2/9 = 11/12.
3.52 A tabular form of the experiment can be established as,
Sample Space x y
HHH
HHT
HTH
THH
HTT
THT
TTH
TTT
3
2
2
2
1
1
1
0
3
1
1
1
−1
−1
−1
−3
So, the joint probability distribution is,
x
f(x, y) 0 1 2 3
−3 1/8
y −1 3/8
1 3/8
3 1/8
Solutions for Exercises in Chapter 3 39
3.53 (a) If (x, y) represents the selection of x kings and y jacks in 3 draws, we must have
x = 0, 1, 2, 3; y = 0, 1, 2, 3; and 0 ≤ x + y ≤ 3. Therefore, (1, 2) represents the
selection of 1 king and 2 jacks which will occur with probability
f(1, 2) =
4
1
4
2
12
3
=
6
55
.
Proceeding in a similar fashion for the other possibilities, we arrive at the following
joint probability distribution:
x
f(x, y) 0 1 2 3
0 1/55 6/55 6/55 1/55
y 1 6/55 16/55 6/55
2 6/55 6/55
3 1/55
(b) P[(X, Y ) ∈ A] = P(X +Y ≥ 2) = 1−P(X +Y < 2) = 1−1/55−6/55−6/55 =
42/55.
3.54 (a) P(H) = 0.4, P(T) = 0.6, and S = {HH,HT, TH, TT}. Let (W, Z) represent
a typical outcome of the experiment. The particular outcome (1, 0) indicating a
total of 1 head and no heads on the first toss corresponds to the event TH. Therefore,
f(1, 0) = P(W = 1, Z = 0) = P(TH) = P(T)P(H) = (0.6)(0.4) = 0.24.
Similar calculations for the outcomes (0, 0), (1, 1), and (2, 1) lead to the following
joint probability distribution:
w
f(w, z) 0 1 2
z 0 0.36 0.24
1 0.24 0.16
(b) Summing the columns, the marginal distribution of W is
w 0 1 2
g(w) 0.36 0.48 0.16
(c) Summing the rows, the marginal distribution of Z is
z 0 1
h(z) 0.60 0.40
(d) P(W ≥ 1) = f(1, 0) + f(1, 1) + f(2, 1) = 0.24 + 0.24 + 0.16 = 0.64.
40 Chapter 3 Random Variables and Probability Distributions
3.55 g(x) = 1
8
R 4
2 (6 − x − y) dy = 3−x
4 , for 0 < x < 2.
So, f(y|x) = f(x,y)
g(x) = 6−x−y
2(3−x) , for 2 < y < 4,
and P(1 < Y < 3 | X = 1) = 1
4
R 3
2 (5 − y) dy = 5
8 .
3.56 Since f(1, 1) 6= g(1)h(1), the variables are not independent.
3.57 X and Y are independent since f(x, y) = g(x)h(y) for all (x, y).
3.58 (a) h(y) = 6
R 1−y
0 x dx = 3(1 − y)2, for 0 < y < 1. Since f(x|y) = f(x,y)
h(y) = 2x
(1−y)2 , for
0 < x < 1 − y, involves the variable y, X and Y are not independent.
(b) P(X > 0.3 | Y = 0.5) = 8
R 0.5
0.3 x dx = 0.64.
3.59 (a) 1 = k
R 1
0
R 1
0
R 2
0 xy2z dx dy dz = 2k
R 1
0
R 1
0 y2z dy dz = 2k
3
R 1
0 z dz = k
3 . So, k = 3.
(b) P
X < 1
4 , Y > 1
2 , 1 < Z < 2
= 3
R 1/4
0
R 1
1/2
R 2
1 xy2z dx dy dz = 9
2
R 1/4
0
R 1
1/2 y2z dy dz
= 21
16
R 1/4
0 z dz = 21
512 .
3.60 g(x) = 4
R 1
0 xy dy = 2x, for 0 < x < 1; h(y) = 4
R 1
0 xy dx = 2y, for 0 < y < 1. Since
f(x, y) = g(x)h(y) for all (x, y), X and Y are independent.
3.61 g(x) = k
R 50
30 (x2 + y2) dy = k
x2y + y3
3
50
30
= k
20x2 + 98,000
3
, and
h(y) = k
20y2 + 98,000
3
.
Since f(x, y) 6= g(x)h(y), X and Y are not independent.
3.62 (a) g(y, z) = 4
9
R 1
0 xyz2 dx = 2
9yz2, for 0 < y < 1 and 0 < z < 3.
(b) h(y) = 2
9
R 3
0 yz2 dz = 2y, for 0 < y < 1.
(c) P
1
4 < X < 1
2 , Y > 1
3 , Z < 2
= 4
9
R 2
1
R 1
1/3
R 1/2
1/4 xyz2 dx dy dz = 7
162 .
(d) Since f(x|y, z) = f(x,y,z)
g(y,z) = 2x, for 0 < x < 1, P
0 < X < 1
2 | Y = 1
4 , Z = 2
=
2
R 1/2
0 x dx = 1
4 .
3.63 g(x) = 24
R 1−x
0 xy dy = 12x(1 − x)2, for 0 < x < 1.
(a) P(X ≥ 0.5) = 12
R 1
0.5 x(1 − x)2 dx =
R 1
0.5(12x − 24x2 + 12x3) dx = 5
16 = 0.3125.
(b) h(y) = 24
R 1−y
0 xy dx = 12y(1 − y)2, for 0 < y < 1.
(c) f(x|y) = f(x,y)
h(y) = 24xy
12y(1−y)2 = 2x
(1−y)2 , for 0 < x < 1 − y.
So, P
X < 1
8 | Y = 3
4
=
R 1/8
0
2x
1/16 dx = 32
R 1/8
0 = 0.25.
3.64 (a)
x 1 3 5 7
f(x) 0.4 0.2 0.2 0.2
(b) P(4 < X ≤ 7) = P(X ≤ 7) − P(X ≤ 4) = F(7) − F(4) = 1 − 0.6 = 0.4.
Solutions for Exercises in Chapter 3 41
3.65 (a) g(x) =
R
∞
0 ye−y(1+x) dy = − 1
1+xye−y(1+x)
∞0
+ 1
1+x
R
∞
0 e−y(1+x) dy
= − 1
(1+x)2 e−y(1+x)
∞0
= 1
(1+x)2 , for x > 0.
h(y) = ye−y
R
∞
0 e−yx dx = −e−y e−yx|∞0 = e−y, for y > 0.
(b) P(X ≥ 2, Y ≥ 2) =
R
∞
2
R
∞
2 ye−y(1+x) dx dy = −
R
∞
2 e−y e−yx|∞2 dy =
R
∞
2 e−3ydy
= − 1
3e−3y
∞2
= 1
3e6 .
3.66 (a) P
X ≤ 1
2 , Y ≤ 1
2
= 3
2
R 1/2
0
R 1/2
0 (x2 + y2) dxdy = 3
2
R 1/2
0
x2y + y3
3
1/2
0
dx
= 3
4
R 1/2
0
x2 + 1
12
dx = 1
16 .
(b) P
X ≥ 3
4
= 3
2
R 1
3/4
x2 + 1
3
dx = 53
128 .
3.67 (a)
x 0 1 2 3 4 5 6
f(x) 0.1353 0.2707 0.2707 0.1804 0.0902 0.0361 0.0120
(b) A histogram is shown next.
1 2 3 4 5 6 7
x
f(x)
0.0
0.1
0.2
0.3
(c)
x 0 1 2 3 4 5 6
F(x) 0.1353 0.4060 0.6767 0.8571 0.9473 0.9834 0.9954
3.68 (a) g(x) =
R 1
0 (x + y) dy = x + 1
2 , for 0 < x < 1, and h(y) = y + 1
2 for 0 < y < 1.
(b) P(X > 0.5, Y > 0.5) =
R 1
0.5
R 1
0.5(x + y) dx dy =
R 1
0.5
x2
2 + xy
1
0.5
dy
=
R 1
0.5
1
2 + y
−
1
8 + y
2
dy = 3
8 .
3.69 f(x) =
5
x
(0.1)x(1 − 0.1)5−x, for x = 0, 1, 2, 3, 4, 5.
3.70 (a) g(x) =
R 2
1
3x−y
9
dy = 3xy−y2/2
9
2
1
= x
3 − 1
6 , for 1 < x < 3, and
h(y) =
R 3
1
3x−y
9
dx = 4
3 − 2
9y, for 1 < y < 2.
(b) No, since g(x)h(y) 6= f(x, y).
(c) P(X > 2) =
R 3
2
x
3 − 1
6
dx =
x2
6 − x
6
3
2
= 2
3 .
42 Chapter 3 Random Variables and Probability Distributions
3.71 (a) f(x) = d
dxF(x) = 1
50e−x/50, for x > 0.
(b) P(X > 70) = 1 − P(X ≤ 70) = 1 − F(70) = 1 − (1 − e−70/50) = 0.2466.
3.72 (a) f(x) = 1
10 , for x = 1, 2, . . . , 10.
(b) A c.d.f. plot is shown next.
F(x)
x
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1 2 3 4 5 6 7 8 9 10
3.73 P(X ≥ 3) = 1
2
R
∞
3 e−y/2 = e−3/2 = 0.2231.
3.74 (a) f(x) ≥ 0 and
R 10
0
1
10 dx = 1. This is a continuous uniform distribution.
(b) P(X ≤ 7) = 1
10
R 7
0 dx = 0.7.
3.75 (a) f(y) ≥ 0 and
R 1
0 f(y) dy = 10
R 1
0 (1 − y)9 dy = − 10
10 (1 − y)10
1
0 = 1.
(b) P(Y > 0.6) =
R 1
0.6 f(y) dy = − (1 − y)10|1
0.6 = (1 − 0.6)10 = 0.0001.
3.76 (a) P(Z > 20) = 1
10
R
∞
20 e−z/10 dz = − e−z/10
∞ 2
0
=
e−2
0
/
1
0
=
0.1353.
(b) P(Z ≤ 10) = − e−z/10
1
0
0 = 1 − e−10/10 = 0.6321.
3.77 (a) g(x1) =
R 1
x1
2 dx2 = 2(1 − x1), for 0 < x1 < 1.
(b) h(x2) =
R x2
0 2 dx1 = 2x2, for 0 < x2 < 1.
(c) P(X1 < 0.2,X2 > 0, 5) =
R 1
0.5
R 0.2
0 2 dx1 dx2 = 2(1 − 0.5)(0.2 − 0) = 0.2.
(d) fX1|X2(x1|x2) = f(x1,x2)
h(x2) = 2
2x2
= 1
x2
, for 0 < x1 < x2.
3.78 (a) fX1(x1) =
R x1
0 6x2 dx2 = 3x2
1, for 0 < x1 < 1. Apparently, fX1R (x1) ≥ 0 and 1
0 fX1(x1) dx1 =
R 1
0 3x2
1 dx1 = 1. So, fX1(x1) is a density function.
(b) fX2|X1(x2|x1) = f(x1,x2)
fX1 (x1) = 6x2
3x21
= 2x2
x21
, for 0 < x2 < x1.
So, P(X2 < 0.5 | X1 = 0.7) = 2
0.72
R 0.5
0 x2 dx2 = 25
49 .
Solutions for Exercises in Chapter 3 43
3.79 (a) g(x) = 9
(16)4y
∞P
x=0
1
4x = 9
(16)4y
1
1−1/4 = 3
4 · 1
4x , for x = 0, 1, 2, . . . ; similarly, h(y) = 3
4 · 1
4y ,
for y = 0, 1, 2, . . . . Since f(x, y) = g(x)h(y), X and Y are independent.
(b) P(X +Y < 4) = f(0, 0)+f(0, 1)+f(0, 2)+f(0, 3)+f(1, 0)+f(1, 1)+f(1, 2)+
f(2, 0) + f(2, 1) + f(3, 0) = 9
16
1 + 1
4 + 1
42 + 1
43 + 1
4 + 1
42 + 1
43 + 1
r2 + 1
43 + 1
43
=
9
16
1 + 2
4 + 3
42 + 4
43
= 63
64 .
3.80 P(the system works) = P(all components work) = (0.95)(0.99)(0.92) = 0.86526.
3.81 P(the system does not fail) = P(at least one of the components works)
= 1−P(all components fail) = 1−(1−0.95)(1−0.94)(1−0.90)(1−0.97) = 0.999991.
3.82 Denote by X the number of components (out of 5) work.
Then, P(the system is operational) = P(X ≥ 3) = P(X = 3) + P(X = 4) + P(X =
5) =
5
3
(0.92)3(1 − 0.92)2 +
5
4
(0.92)4(1 − 0.92) +
5
5
(0.92)5 = 0.9955.
Chapter 4
Mathematical Expectation
4.1 E(X) = 1
a2
R a
−a
R√a2−y2
−√a2−y2
x dx dy = 1
a2
h
a2−y2
2
−
a2−y2
2
i
dy = 0.
4.2 E(X) =
P3
x=0
x f(x) = (0)(27/64) + (1)(27/64) + (2)(9/64) + (3)(1/64) = 3/4.
4.3 μ = E(X) = (20)(1/5) + (25)(3/5) + (30)(1/5) = 25 cents.
4.4 Assigning wrights of 3w and w for a head and tail, respectively. We obtain P(H) = 3/4
and P(T) = 1/4. The sample space for the experiment is S = {HH,HT, TH, TT}.
Now if X represents the number of tails that occur in two tosses of the coin, we have
P(X = 0) = P(HH) = (3/4)(3/4) = 9/16,
P(X = 1) = P(HT) + P(TH) = (2)(3/4)(1/4) = 3/8,
P(X = 2) = P(TT) = (1/4)(1/4) = 1/16.
The probability distribution for X is then
x 0 1 2
f(x) 9/16 3/8 1/16
from which we get μ = E(X) = (0)(9/16) + (1)(3/8) + (2)(1/16) = 1/2.
4.5 μ = E(X) = (0)(0.41) + (1)(0.37) + (2)(0.16) + (3)(0.05) + (4)(0.01) = 0.88.
4.6 μ = E(X) = ($7)(1/12)+($9)(1/12)+($11)(1/4)+($13)(1/4)+($15)(1/6)+($17)(1/6)
= $12.67.
4.7 Expected gain = E(X) = (4000)(0.3) + (−1000)(0.7) = $500.
4.8 Let X = profit. Then
μ = E(X) = (250)(0.22) + (150)(0.36) + (0)(0.28) + (−150)(0.14) = $88.
45
46 Chapter 4 Mathematical Expectation
4.9 Let c = amount to play the game and Y = amount won.
y 5 − c 3 − c −c
f(y) 2/13 2/13 9/13
E(Y ) = (5 − c)(2/13) + (3 − c)(2/13) + (−c)(9/13) = 0. So, 13c = 16 which implies
c = $1.23.
4.10 μX =
P
xg(x) = (1)(0.17) + (2)(0.5) + (3)(0.33) = 2.16,
μY =
P
yh(y) = (1)(0.23) + (2)(0.5) + (3)(0.27) = 2.04.
4.11 For the insurance of $200,000 pilot, the distribution of the claim the insurance company
would have is as follows:
Claim Amount $200,000 $100,000 $50,000 0
f(x) 0.002 0.01 0.1 0.888
So, the expected claim would be
($200, 000)(0.002) + ($100, 000)(0.01) + ($50, 000)(0.1) + ($0)(0.888) = $6, 400.
Hence the insurance company should charge a premium of $6, 400 + $500 = $6, 900.
4.12 E(X) =
R 1
0 2x(1 − x) dx = 1/3. So, (1/3)($5, 000) = $1, 667.67.
4.13 E(X) = 4
R 1
0
x
1+x2 dx = ln 4
.
4.14 E(X) =
R 1
0
2x(x+2)
5 dx = 8
15 .
4.15 E(X) =
R 1
0 x2 dx +
R 2
1 x(2 − x) dx = 1. Therefore, the average number of hours per
year is (1)(100) = 100 hours.
4.16 P(X1 + X2 = 1) = P(X1 = 1,X2 = 0) + P(X1 = 0,X2 = 1)
=
(980
1 )(20
1 )
(1000
2 )
+
(980
1 )(20
1 )
(1000
2 )
= (2)(0.0392) = 0.0784.
4.17 The probability density function is,
x −3 6 9
f(x) 1/6 1/2 1/3
g(x) 25 169 361
μg(X) = E[(2X + 1)2] = (25)(1/6) + (169)(1/2) + (361)(1/3) = 209.
4.18 E(X2) = (0)(27/64) + (1)(27/64) + (4)(9/64) + (9)(1/64) = 9/8.
4.19 Let Y = 1200X − 50X2 be the amount spent.
Solutions for Exercises in Chapter 4 47
x 0 1 2 3
f(x) 1/10 3/10 2/5 1/5
y = g(x) 0 1150 2200 3150
μY = E(1200X − 50X2) = (0)(1/10) + (1150)(3/10) + (2200)(2/5) + (3150)(1/5)
= $1, 855.
4.20 E[g(X)] = E(e2X/3) =
R
∞
0 e2x/3e−x dx =
R
∞
0 e−x/3 dx = 3.
4.21 E(X2) =
R 1
0 2x2(1 − x) dx = 1
6 . Therefore, the average profit per new automobile is
(1/6)($5000.00) = $833.33.
4.22 E(Y ) = E(X + 4) =
R
∞
0 32(x + 4) 1
(x+4)3 dx = 8 days.
4.23 (a) E[g(X, Y )] = E(XY 2) =
P
x
P
y
xy2f(x, y)
= (2)(1)2(0.10) + (2)(3)2(0.20) + (2)(5)2(0.10) + (4)(1)2(0.15) + (4)(3)2(0.30)
+ (4)(5)2(0.15) = 35.2.
(b) μX = E(X) = (2)(0.40) + (4)(0.60) = 3.20,
μY = E(Y ) = (1)(0.25) + (3)(0.50) + (5)(0.25) = 3.00.
4.24 (a) E(X2Y − 2XY ) =
P3
x=0
P2
y=0
(x2y − 2xy)f(x, y) = (1− 2)(18/70)+ (4− 4)(18/70)+
· · · + (8 − 8)(3/70) = −3/7.
(b)
x 0 1 2 3
g(x) 5/70 30/70 30/70 5/70
y 0 1 2
h(y) 15/70 40/70 15/70
μX = E(X) = (0)(5/70) + (1)(30/70) + (2)(30/70) + (3)(5/70) = 3/2,
μY = E(Y ) = (0)(15/70) + (1)(40/70) + (2)(15/70) = 1.
4.25 μX+Y = E(X +Y ) =
P3
x=0
P3
y=0
(x+y)f(x, y) = (0+0)(1/55)+(1+0)(6/55)+· · ·+(0+
3)(1/55) = 2.
4.26 E(Z) = E(√X2 + Y 2) =
R 1
0
R 1
0 4xy
p
x2 + y2 dx dy = 4
3
R 1
0 [y(1 + y2)3/2 − y4] dy
= 8(23/2 − 1)/15 = 0.9752.
4.27 E(X) = 1
2000
R
∞
0 x exp(−x/2000) dx = 2000
R
∞
0 y exp(−y) dy = 2000.
4.28 (a) The density function is shown next.
f(x)
23.75 26.25
2/5
48 Chapter 4 Mathematical Expectation
(b) E(X) = 2
5
R 26.25
23.75 x dx = 1
5 (26.252 − 23.752) = 25.
(c) The mean is exactly in the middle of the interval. This should not be surprised
due to the symmetry of the density at 25.
4.29 (a) The density function is shown next
f(x)
0 1 1.5 2 2.5 3 3.5 4
1
2
3
(b) μ = E(X) =
R
∞
1 3x−3 dx = 3
2 .
4.30 E(Y ) = 1
4
R
∞
0 ye−y/4 dy = 4.
4.31 (a) μ = E(Y ) = 5
R 1
0 y(1 − y)4 dy = −
R 1
0 y d(1 − y)5 =
R
∞
0 (1 − y)5 dy = 1
6 .
(b) P(Y > 1/6) =
R 1
1/6 5(1 − y)4 dy = − (1 − y)5|1
1/6 = (1 − 1/6)5 = 0.4019.
4.32 (a) A histogram is shown next.
1 2 3 4 5
x
f(x)
0
0.1
0.2
0.3
0.4
(b) μ = (0)(0.41) + (1)(0.37) + (2)(0.16) + (3)(0.05) + (4)(0.01) = 0.88.
(c) E(X2) = (0)2(0.41) + (1)2(0.37) + (2)2(0.16) + (3)2(0.05) + (4)2(0.01) = 1.62.
(d) V ar(X) = 1.62 − 0.882 = 0.8456.
4.33 μ = $500. So, σ2 = E[(X − μ)2] =
P
x
(x − μ)2f(x) = (−1500)2(0.7) + (3500)2(0.3) =
$5, 250, 000.
Solutions for Exercises in Chapter 4 49
4.34 μ = (−2)(0.3) + (3)(0.2) + (5)(0.5) = 2.5 and
E(X2) = (−2)2(0.3) + (3)2(0.2) + (5)2(0.5) = 15.5.
So, σ2 = E(X2) − μ2 = 9.25 and σ = 3.041.
4.35 μ = (2)(0.01) + (3)(0.25) + (4)(0.4) + (5)(0.3) + (6)(0.04) = 4.11,
E(X2) = (2)2(0.01) + (3)2(0.25) + (4)2(0.4) + (5)2(0.3) + (6)2(0.04) = 17.63.
So, σ2 = 17.63 − 4.112 = 0.74.
4.36 μ = (0)(0.4) + (1)(0.3) + (2)(0.2) + (3)(0.1) = 1.0,
and E(X2) = (0)2(0.4) + (1)2(0.3) + (2)2(0.2) + (3)2(0.1) = 2.0.
So, σ2 = 2.0 − 1.02 = 1.0.
4.37 It is know μ = 1/3.
So, E(X2) =
R 1
0 2x2(1 − x) dx = 1/6 and σ2 = 1/6 − (1/3)2 = 1/18. So, in the actual
profit, the variance is 1
18 (5000)2.
4.38 It is known μ = 8/15.
Since E(X2) =
R 1
0
2
5x2(x + 2) dx = 11
30 , then σ2 = 11/30 − (8/15)2 = 37/450.
4.39 It is known μ = 1.
Since E(X2) =
R 1
0 x2 dx +
R 2
1 x2(2 − x) dx = 7/6, then σ2 = 7/6 − (1)2 = 1/6.
4.40 μg(X) = E[g(X)] =
R 1
0 (3x2 + 4)
2x+4
5
dx = 1
5
R 1
0 (6x3 + 12x2 + 8x + 16) dx = 5.1.
So, σ2 = E[g(X) − μ]2 =
R 1
0 (3x2 + 4 − 5.1)2
2x+4
5
dx
=
R 1
0 (9x4 − 6.6x2 + 1.21)
2x+4
5
dx = 0.83.
4.41 It is known μg(X) = E[(2X + 1)2] = 209. Hence
σ2
g(X) =
P
x
[(2X + 1)2 − 209]2g(x)
= (25 − 209)2(1/6) + (169 − 209)2(1/2) + (361 − 209)2(1/3) = 14, 144.
So, σg(X) = √14, 144 = 118.9.
4.42 It is known μg(X) = E(X2) = 1/6. Hence
σ2
g(X) =
R 1
0 2
x2 − 1
6
2
(1 − x) dx = 7/180.
4.43 μY = E(3X − 2) = 1
4
R
∞
0 (3x − 2)e−x/4 dx = 10. So
σ2
Y = E{[(3X − 2) − 10]2} = 9
4
R
∞
0 (x − 4)2e−x/4 dx = 144.
4.44 E(XY ) =
P
x
P
y
xyf(x, y) = (1)(1)(18/70) + (2)(1)(18/70)
+ (3)(1)(2/70) + (1)(2)(9/70) + (2)(2)(3/70) = 9/7;
μX =
P
x
P
y
xf(x, y) = (0)f(0, 1) + (0)f(0, 2) + (1)f(1, 0) + · · · + (3)f(3, 1) = 3/2,
and μY = 1.
So, σXY = E(XY ) − μXμY = 9/7 − (3/2)(1) = −3/14.
50 Chapter 4 Mathematical Expectation
4.45 μX =
P
x
xg(x) = 2.45, μY =
P
y
yh(y) = 2.10, and
E(XY ) =
P
x
P
x
xyf(x, y) = (1)(0.05) + (2)(0.05) + (3)(0.10) + (2)(0.05)
+ (4)(0.10) + (6)(0.35) + (3)(0) + (6)(0.20) + (9)(0.10) = 5.15.
So, σXY = 5.15 − (2.45)(2.10) = 0.005.
4.46 From previous exercise, k =
3
392
10−4, and g(x) = k
20x2 + 98000
3
, with
μX = E(X) =
R 50
30 xg(x) dx = k
R 50
30
20x3 + 98000
3 x
dx = 40.8163.
Similarly, μY = 40.8163. On the other hand,
E(XY ) = k
R 50
30
R 50
30 xy(x2 + y2) dy dx = 1665.3061.
Hence, σXY = E(XY ) − μXμY = 1665.3061 − (40.8163)2 = −0.6642.
4.47 g(x) = 2
3
R 1
0 (x + 2y) dy = 2
3 (x + 1, for 0 < x < 1, so μX = 2
3
R 1
0 x(x + 1) dx = 5
9 ;
h(y) = 2
3
R 1
0 (x + 2y) dx = 2
3
1
2 + 2y
, so μY = 2
3
R 1
0 y
1
2 + 2y
dy = 11
18 ; and
E(XY ) = 2
3
R 1
0
R 1
0 xy(x + 2y) dy dx = 1
3 .
So, σXY = E(XY ) − μXμY = 1
3 −
5
9
11
18
= −0.0062.
4.48 Since σXY = Cov(a + bX,X) = bσ2
X and σ2
Y = b2σ2
X , then
ρ = XY
X
σY = b 2
X √ 2
X b2 2
X
= b
|b|
= sign of b.
Hence ρ = 1 if b > 0 and ρ = −1 if b < 0.
4.49 E(X) = (0)(0.41) + (1)(0.37) + (2)(0.16) + (3)(0.05) + (4)(0.01) = 0.88
and E(X2) = (0)2(0.41) + (1)2(0.37) + (2)2(0.16) + (3)2(0.05) + (4)2(0.01) = 1.62.
So, V ar(X) = 1.62 − 0.882 = 0.8456 and σ = √0.8456 = 0.9196.
4.50 E(X) = 2
R 1
0 x(1 − x) dx = 2
x2
2 − x3
3
1
0
= 1
3 and
E(X2) = 2
R 1
0 x2(1 − x) dx = 2
x3
3 − x4
4
1
0
= 1
6 . Hence,
V ar(X) = 1
6 −
1
3
2 = 1
18 , and σ =
p
1/18 = 0.2357.
4.51 Previously we found μ = 4.11 and σ2 = 0.74, Therefore,
μg(X) = E(3X − 2) = 3μ − 2 = (3)(4.11) − 2 = 10.33 and σg(X) = 9σ2 = 6.66.
4.52 Previously we found μ = 1 and σ2 = 1. Therefore,
μg(X) = E(5X + 3) = 5μ + 3 = (5)(1) + 3 = 8 and σg(X) = 25σ2 = 25.
4.53 Let X = number of cartons sold and Y = profit.
We can write Y = 1.65X + (0.90)(5 − X) − 6 = 0.75X − 1.50. Now
E(X) = (0)(1/15)+(1)(2/15)+(2)(2/15)+(3)(3/15)+(4)(4/15)+(5)(3/15) = 46/15,
and E(Y ) = (0.75)E(X) − 1.50 = (0.75)(46/15) − 1.50 = $0.80.
4.54 μX = E(X) = 1
4
R
∞
0 xe−x/4 dx = 4.
Therefore, μY = E(3X − 2) = 3E(X) − 2 = (3)(4) − 2 = 10.
Since E(X2) = 1
4
R
∞
0 x2e−x/4 dx = 32, therefore, σ2
X = E(X2) − μ2
X = 32 − 16 = 16.
Hence σ2
Y = 9σ2
X = (9)(16) = 144.
Solutions for Exercises in Chapter 4 51
4.55 E(X) = (−3)(1/6) + (6)(1/2) + (9)(1/3) = 11/2,
E(X2) = (−3)2(1/6) + (6)2(1/2) + (9)2(1/3) = 93/2. So,
E[(2X + 1)2] = 4E(X2) + 4E(X) + 1 = (4)(93/2) + (4)(11/2) + 1 = 209.
4.56 Since E(X) =
R 1
0 x2 dx +
R 2
1 x(2 − x) dx = 1, and
E(X2) =
R 1
0 x32 dx +
R 2
1 x2(2 − x) dx = 7/6,then
E(Y ) = 60E(X2) + 39E(X) = (60)(7/6) + (39)(1) = 109 kilowatt hours.
4.57 The equations E[(X − 1)2] = 10 and E[(X − 2)2] = 6 may be written in the form:
E(X2) − 2E(X) = 9, E(X2) − 4E(X) = 2.
Solving these two equations simultaneously we obtain
E(X) = 7/2, and E(X2) = 16.
Hence μ = 7/2 and σ2 = 16 − (7/2)2 = 15/4.
4.58 E(X) = (2)(0.40) + (4)(0.60) = 3.20, and
E(Y ) = (1)(0.25) + (3)(0.50) + (5)(0.25) = 3. So,
(a) E(2X − 3Y ) = 2E(X) − 3E(Y ) = (2)(3.20) − (3)(3.00) = −2.60.
(b) E(XY ) = E(X)E(Y ) = (3.20)(3.00) = 9.60.
4.59 E(2XY 2 − X2Y ) = 2E(XY 2) − E(X2Y ). Now,
E(XY 2) =
P2
x=0
P2
y=0
xy2f(x, y) = (1)(1)2(3/14) = 3.14, and
E(X2Y ) =
P2
x=0
P2
y=0
x2yf(x, y) = (1)2(1)(3/14) = 3.14.
Therefore, E(2XY 2 − X2Y ) = (2)(3/14) − (3/14) = 3/14.
4.60 Using μ = 60 and σ = 6 and Chebyshev’s theorem
P(μ − kσ < X < μ + kσ) ≥ 1 −
1
k2 ,
since from μ + kσ = 84 we obtain k = 4.
So, P(X < 84) ≥ P(36 < X < 84) ≥ 1 − 1
42 = 0.9375. Therefore,
P(X ≥ 84) ≤ 1 − 0.9375 = 0.0625.
Since 1000(0.0625) = 62.5, we claim that at most 63 applicants would have a score as
84 or higher. Since there will be 70 positions, the applicant will have the job.
4.61 μ = 900 hours and σ = 50 hours. Solving μ − kσ = 700 we obtain k = 4.
So, using Chebyshev’s theorem with P(μ − 4σ < X < μ + 4σ) ≥ 1 − 1/42 = 0.9375,
we obtain P(700 < X < 1100) ≥ 0.9375. Therefore, P(X ≤ 700) ≤ 0.03125.
52 Chapter 4 Mathematical Expectation
4.62 μ = 52 and σ = 6.5. Solving μ + kσ = 71.5 we obtain k = 3. So,
P(μ − 3σ < X < μ + 3σ) ≥ 1 −
1
32 = 0.8889,
which is
P(32.5 < X < 71.5) ≥ 0.8889.
we obtain P(X > 71.5) < 1−0.8889
2 = 0.0556 using the symmetry.
4.63 n = 500, μ = 4.5 and σ = 2.8733. Solving μ + k(σ/√500) = 5 we obtain
k =
5 − 4.5
2.87333/√500
=
0.5
0.1284
= 3.8924.
So, P(4 ≤ ¯X ≤ 5) ≥ 1 − 1
k2 = 0.9340.
4.64 σ2
Z = σ2
−2X+4Y −3 = 4σ2
X + 16σ2
Y = (4)(5) + (16)(3) = 68.
4.65 σ2
Z = σ2
−2X+4Y −3 = 4σ2
X + 16σ2
Y − 16σXY = (4)(5) + (16)(3) − (16)(1) = 52.
4.66 (a) P(6 < X < 18) = P[12 − (2)(3) < X < 12 + (2)(3)] ≥ 1 − 1
22 = 3
4 .
(b) P(3 < X < 21) = P[12 − (3)(3) < X < 12 + (3)(3)] ≥ 1 − 1
32 = 8
9 .
4.67 (a) P(|X − 10| ≥ 3) = 1 − P(|X − 10| < 3)
= 1 − P[10 − (3/2)(2) < X < 10 + (3/2)(2)] ≤ 1 −
h
1 − 1
(3/2)2
i
= 4
9 .
(b) P(|X − 10| < 3) = 1 − P(|X − 10| ≥ 3) ≥ 1 − 4
9 = 5
9 .
(c) P(5 < X < 15) = P[10 − (5/2)(2) < X < 10 + (5/2)(2)] ≥ 1 − 1
(5/2)2 = 21
25 .
(d) P(|X − 10| ≥ c) ≤ 0.04 implies that P(|X − 10| < c) ≥ 1 − 0.04 = 0.96.
Solving 0.96 = 1 − 1
k2 we obtain k = 5. So, c = kσ = (5)(2) = 10.
4.68 μ = E(X) = 6
R 1
0 x2(1 − x) dx = 0.5, E(X2) = 6
R 1
0 x3(1 − x) dx = 0.3, which imply
σ2 = 0.3 − (0.5)2 = 0.05 and σ = 0.2236. Hence,
P(μ − 2σ < X < μ + 2σ) = P(0.5 − 0.4472 < X < 0.5 + 0.4472)
= P(0.0528 < X < 0.9472) = 6
Z 0.9472
0.0528
x(1 − x) dx = 0.9839,
compared to a probability of at least 0.75 given by Chebyshev’s theorem.
4.69 It is easy to see that the expectations of X and Y are both 3.5. So,
(a) E(X + Y ) = E(X) + E(Y ) = 3.5 + 3.5 = 7.0.
(b) E(X − Y ) = E(X) − E(Y ) = 0.
Solutions for Exercises in Chapter 4 53
(c) E(XY ) = E(X)E(Y ) = (3.5)(3.5) = 12.25.
4.70 E(Z) = E(XY ) = E(X)E(Y ) =
R 1
0
R
∞
2 16xy(y/x3) dx dy = 8/3.
4.71 E[g(X, Y )] = E(X/Y 3 + X2Y ) = E(X/Y 3) + E(X2Y ).
E(X/Y 3) =
R 2
1
R 1
0
2x(x+2y)
7y3 dx dy = 2
7
R 2
1
1
3y3 + 1
y2
dy = 15
84 ;
E(X2Y ) =
R 2
1
R 1
0
2x2y(x+2y)
7 dx dy = 2
7
R 2
1 y
1
4 + 2y
3
dy = 139
252 .
Hence, E[g(X, Y )] = 15
84 + 139
252 = 46
63 .
4.72 μX = μY = 3.5. σ2
X = σ2
Y = [(1)2 + (2)2 + · · · + (6)2](1/6) − (3.5)2 = 35
12 .
(a) σ2X−Y = 4σ2
X + σ2
Y = 175
12 ;
(b) σX+3Y −5 = σ2
X + 9σ2
Y = 175
6 .
4.73 (a) μ = 1
5
R 5
0 x dx = 2.5, σ2 = E(X2) − μ2 = 1
5x2
R 5
0 x2 dx − 2.52 = 2.08.
So, σ = √σ2 = 1.44.
(b) By Chebyshev’s theorem,
P[2.5 − (2)(1.44) < X < 2.5 + (2)(1.44)] = P(−0.38 < X < 5.38) ≥ 0.75.
Using integration, P(−0.38 < X < 5.38) = 1 ≥ 0.75;
P[2.5 − (3)(1.44) < X < 2.5 + (3)(1.44)] = P(−1.82 < X < 6.82) ≥ 0.89.
Using integration, P(−1.82 < X < 6.82) = 1 ≥ 0.89.
4.74 P = I2R with R = 50, μI = E(I) = 15 and σ2
I = V ar(I) = 0.03.
E(P) = E(I2R) = 50E(I2) = 50[V ar(I) + μ2
I ] = 50(0.03 + 152) = 11251.5. If we use
the approximation formula, with g(I) = I2, g′(I) = 2I and g′′(I) = 2, we obtain,
E(P) ≈ 50
g(μI) + 2
σ2
I
2
= 50(152 + 0.03) = 11251.5.
Since V ar[g(I)] ≈
h
@g(i)
@i
i2
i=μI
σ2
I , we obtain
V ar(P) = 502V ar(I2) = 502(2μI)2σ2
I = 502(30)2(0.03) = 67500.
4.75 For 0 < a < 1, since g(a) =
∞P
x=0
ax = 1
1−a , g′(a) =
∞P
x=1
xax−1 = 1
(1−a)2 and
g′′(a) =
∞P
x=2
x(x − 1)ax−2 = 2
(1−a)3 .
54 Chapter 4 Mathematical Expectation
(a) E(X) = (3/4)
∞P
x=1
x(1/4)x = (3/4)(1/4)
∞P
x=1
x(1/4)x−1 = (3/16)[1/(1 − 1/4)2]
= 1/3, and E(Y ) = E(X) = 1/3.
E(X2) − E(X) = E[X(X − 1)] = (3/4)
∞P
x=2
x(x − 1)(1/4)x
= (3/4)(1/4)2 ∞P
x=2
x(x − 1)(1/4)x−2 = (3/43)[2/(1 − 1/4)3] = 2/9.
So, V ar(X) = E(X2) − [E(X)]2 = [E(X2) − E(X)] + E(X) − [E(X)]2
2/9 + 1/3 − (1/3)2 = 4/9, and V ar(Y ) = 4/9.
(b) E(Z) = E(X) + E(Y ) = (1/3) + (1/3) = 2/3, and
V ar(Z) = V ar(X +Y ) = V ar(X)+V ar(Y ) = (4/9)+ (4/9) = 8/9, since X and
Y are independent (from Exercise 3.79).
4.76 (a) g(x) = 3
2
R 1
0 (x2 + y2) dy = 1
2 (3x2 + 1) for 0 < x < 1 and
h(y) = 1
2 (3y2 + 1) for 0 < y < 1.
Since f(x, y) 6= g(x)h(y), X and Y are not independent.
(b) E(X + Y ) = E(X) + E(Y ) = 2E(X) =
R 1
0 x(3x2 + 1) dx = 3/4 + 1/2 = 5/4.
E(XY ) = 3
2
R 1
0
R 1
0 xy(x2 + y2) dx dy = 3
2
R 1
0 y
1
4 + y2
2
dy
= 3
2
1
4
1
2
+
1
2
1
4
= 3
8 .
(c) V ar(X) = E(X2) − [E(X)]2 = 1
2
R 1
0 x2(3x2 + 1) dx −
5
8
2 = 7
15 − 25
64 = 73
960 , and
V ar(Y ) = 73
960 . Also, Cov(X, Y ) = E(XY ) − E(X)E(Y ) = 3
8 −
5
8
2
= − 1
64 .
(d) V ar(X + Y ) = V ar(X) + V ar(Y ) + 2Cov(X, Y ) = 2 73
960 − 2 1
64 = 29
240 .
4.77 (a) E(Y ) =
R
∞
0 ye−y/4 dy = 4.
(b) E(Y 2) =
R
∞
0 y2e−y/4 dy = 32 and V ar(Y ) = 32 − 42 = 16.
4.78 (a) The density function is shown next.
x
f(x)
0 7 8
1
(b) E(Y ) =
R 8
7 y dy = 1
2 [82 − 72] = 15
2 = 7.5,
E(Y 2) =
R 8
7 y2 dy = 1
3 [83 − 73] = 169
3 , and V ar(Y ) = 169
3 −
15
2
2
= 1
12 .
4.79 Using the exact formula, we have
E(eY ) =
Z 8
7
ey dy = ey|8
7 = 1884.32.
Solutions for Exercises in Chapter 4 55
Using the approximation, since g(y) = ey, so g′′(y) = ey. Hence, using the approximation
formula,
E(eY ) ≈ eμY + eμY σ2
Y
2
=
1 +
1
24
e7.5 = 1883.38.
The approximation is very close to the true value.
4.80 Using the exact formula, E(Z2) =
R 8
7 e2y dy = 1
2 e2y|8
7 = 3841753.12. Hence,
V ar(Z) = E(Z2) − [E(Z)]2 = 291091.3.
Using the approximation formula, we have
V ar(eY ) = (eμY )2V ar(Y ) =
e(2)(7.5)
12
= 272418.11.
The approximation is not so close to each other. One reason is that the first order
approximation may not always be good enough.
4.81 Define I1 = {xi| |xi − μ| < kσ} and I2 = {xi| |xi − μ| ≥ kσ}. Then
σ2 = E[(X − μ)2] =
X
x
(x − μ)2f(x) =
X
xi∈I1
(xi − μ)2f(xi) +
X
xi∈I2
(xi − μ)2f(xi)
≥
X
xi∈I2
(xi − μ)2f(xi) ≥ k2σ2
X
xi∈I2
f(xi) = k2σ2P(|X − μ| ≥ kσ),
which implies
P(|X − μ| ≥ kσ) ≤
1
k2 .
Hence, P(|X − μ| < kσ) ≥ 1 − 1
k2 .
4.82 E(XY ) =
R 1
0
R 1
0 xy(x+y) dx dy = 1
3 , E(X) =
R 1
0
R 1
0 x(x+y) dx dy = 7
12 and E(Y ) = 7
12 .
Therefore, σXY = E(XY ) − μXμY = 1
3 −
7
12
2
= − 1
144 .
4.83 E(Y − X) =
R 1
0
R y
0 2(y − x) dx dy =
R 1
0 y2 dy = 1
3 . Therefore, the average amount of
kerosene left in the tank at the end of each day is (1/3)(1000) = 333 liters.
4.84 (a) E(X) =
R
∞
0
x
5 e−x/5 dx = 5.
(b) E(X2) =
R
∞
0
x2
5 e−x/5 dx = 50, so V ar(X) = 50 − 52 = 25, and σ = 5.
(c) E[(X + 5)2] = E{[(X − 5) + 10]2} = E[(X − 5)2] + 102 + 20E(X − 5)
= V ar(X) + 100 = 125.
4.85 E(XY ) = 24
R 1
0
R 1−y
0 x2y2 dx dy = 8
R 1
0 y2(1 − y)3 dy = 2
15 ,
μX = 24
R 1
0
R 1−y
0 x2y dx dy = 2
5 and μY = 24
R 1
0
R 1−y
0 xy2 dx dy = 2
5 . Therefore,
σXY = E(XY ) − μXμY = 2
15 −
2
5
2
= − 2
75 .
56 Chapter 4 Mathematical Expectation
4.86 E(X + Y ) =
R 1
0
R 1−y
0 24(x + y)xy dx dy = 4
5 .
4.87 (a) E(X) =
R
∞
0
x
900e−x/900 dx = 900 hours.
(b) E(X2) =
R
∞
0
x2
900e−x/900 dx = 1620000 hours2.
(c) V ar(X) = E(X2) − [E(X)]2 = 810000 hours2 and σ = 900 hours.
4.88 It is known g(x) = 2
3 (x + 1), for 0 < x < 1, and h(y) = 1
3 (1 + 4y), for 0 < y < 1.
(a) μX =
R 1
0
2
3x(x + 1) dx = 5
9 and μY =
R 1
0
1
3y(1 + 4y) dy = 11
18 .
(b) E[(X + Y )/2] = 1
2 [E(X) + E(Y )] = 7
12 .
4.89 Cov(aX, bY ) = E[(aX − aμX)(bY − bμY )] = abE[(X − μX)(Y − μY )] = abCov(X, Y ).
4.90 It is known μ = 900 and σ = 900. For k = 2,
P(μ − 2σ < X < μ + 2σ) = P(−900 < X < 2700) ≥ 0.75
using Chebyshev’s theorem. On the other hand,
P(μ − 2σ < X < μ + 2σ) = P(−900 < X < 2700) = 1 − e−3 = 0.9502.
For k = 3, Chebyshev’s theorem yields
P(μ − 3σ < X < μ + 3σ) = P(−1800 < X < 3600) ≥ 0.8889,
while P(−1800 < X < 3600) = 1 − e−4 = 0.9817.
4.91 g(x) =
R 1
0
16y
x3 dy = 8
x3 , for x > 2, with μX =
R
∞
2
8
x2 dx = − 8
x
∞2
= 4,
h(y) =
R
∞
2
16y
x3 dx = − 8y
x2
∞2
= 2y, for 0 < y < 1, with μY =
R 1
0 2y2 = 2
3 , and
E(XY ) =
R
∞
2
R 1
0
16y2
x2 dy dx = 16
3
R
∞
2
1
x2 dx = 8
3 . Hence,
σXY = E(XY ) − μXμY = 8
3 − (4)
2
3
= 0.
4.92 Since σXY = 1, σ2
X = 5 and σ2
Y = 3, we have ρ = XY
X Y
= 1 √(5)(3)
= 0.2582.
4.93 (a) From Exercise 4.37, we have σ2 = 1/18, so σ = 0.2357.
(b) Also, μX = 1/3 from Exercise 4.12. So,
P(μ − 2σ < X < μ + 2σ) = P[1/3 − (2)(0.2357) < X < 1/3 + (2)(0.2357)]
= P(0 < X < 0.8047) =
Z 0.8047
0
2(1 − x) dx = 0.9619.
Using Chebyshev’s theorem, the probability of this event should be larger than
0.75, which is true.
(c) P(profit > $500) = P(X > 0.1) =
R 1
0.1 2(1 − x) = 0.81.
Solutions for Exercises in Chapter 4 57
4.94 Since g(0)h(0) = (0.17)(0.23) 6= 0.10 = f(0, 0), X and Y are not independent.
4.95 E(X) = (−5000)(0.2) + (10000)(0.5) + (30000)(0.3) = $13, 000.
4.96 (a) f(x) =
3
x
(0.15)x(0.85)3−x, for x = 0, 1, 2, 3.
x 0 1 2 3
f(x) 0.614125 0.325125 0.057375 0.003375
(b) E(X) = 0.45.
(c) E(X2) = 0.585, so V ar(X) = 0.585 − 0.452 = 0.3825.
(d) P(X ≤ 2) = 1 − P(X = 3) = 1 − 0.003375 = 0.996625.
(e) 0.003375.
(f) Yes.
4.97 (a) E(X) = (−$15k)(0.05)+($15k)(0.15)+($25k)(0.30)+($40k)(0.15)+($50k)(0.10)+
($100k)(0.05) + ($150k)(0.03) + ($200k)(0.02) = $33.5k.
(b) E(X2) = 2, 697, 500, 000 dollars2. So, σ =
p
E(X2) − [E(X)]2 = $39.689k.
4.98 (a) E(X) = 3
4×503
R 50
−50 x(502 − x2) dx = 0.
(b) E(X2) = 3
4×503
R 50
−50 x2(502 − x2) dx = 500.
(c) σ =
p
E(X2) − [E(X)]2 = √500 − 0 = 22.36.
4.99 (a) The marginal density of X is
x1 0 1 2 3 4
fX1(x1) 0.13 0.21 0.31 0.23 0.12
(b) The marginal density of Y is
x2 0 1 2 3 4
fX2(x2) 0.10 0.30 0.39 0.15 0.06
(c) Given X2 = 3, the conditional density function of X1 is f(x1, 3)/0.15. So
x1 0 1 2 3 4
fX2(x2) 7
15
1
5
1
15
1
5
1
15
(d) E(X1) = (0)(0.13) + (1)(0.21) + (2)(0.31) + (3)(0.23) + (4)(0.12) = 2.
(e) E(X2) = (0)(0.10) + (1)(0.30) + (2)(0.39) + (3)(0.15) + (4)(0.06) = 1.77.
(f) E(X1|X2 = 3) = (0)
7
15
+ (1)
1
5
+ (2)
1
15
+ (3)
1
5
+ (4)
1
15
= 18
15 = 6
5 = 1.2.
(g) E(X2
1 ) = (0)2(0.13) + (1)2(0.21) + (2)2(0.31) + (3)2(0.23) + (4)2(0.12) = 5.44.
So, σX1 =
p
E(X2
1 ) − [E(X1)]2 = √5.44 − 22 = √1.44 = 1.2.
4.100 (a) The marginal densities of X and Y are, respectively,
58 Chapter 4 Mathematical Expectation
x 0 1 2
g(x) 0.2 0.32 0.48
y 0 1 2
h(y) 0.26 0.35 0.39
The conditional density of X given Y = 2 is
x 0 1 2
fX|Y =2(x|2) 4
39
5
39
30
39
(b) E(X) = (0)(0.2) + (1)(0.32) + (2)(0.48) = 1.28,
E(X2) = (0)2(0.2) + (1)2(0.32) + (2)2(0.48) = 2.24, and
V ar(X) = 2.24 − 1.282 = 0.6016.
(c) E(X|Y = 2) = (1) 5
39 + (2) 30
39 = 65
39 and E(X2|Y = 2) = (1)2 5
39 + (2)2 30
39 = 125
39 . So,
V ar(X) = 125
39 −
65
39
2 = 650
1521 = 50
117 .
4.101 The profit is 8X + 3Y − 10 for each trip. So, we need to calculate the average of this
quantity. The marginal densities of X and Y are, respectively,
x 0 1 2
g(x) 0.34 0.32 0.34
y 0 1 2 3 4 5
h(y) 0.05 0.18 0.15 0.27 0.19 0.16
So, E(8X+3Y −10) = (8)[(1)(0.32)+(2)(0.34)]+(3)[(1)(0.18)+(2)(0.15)+(3)(0.27)+
(4)(0.19) + (5)(0.16)] − 10 = $6.55.
4.102 Using the approximation formula, V ar(Y ) ≈
Pk
i=1
h
@h(x1,x2,...,xk)
@xi
i2
xi=μi, 1≤i≤k
σ2
i , we
have
V ar( ˆ Y ) ≈
X2
i=0
∂eb0+b1k1+b2k2
∂bi
2
bi= i, 0≤i≤2
σ2
bi = e2( 0+k1 1+k2 2)(σ2
0 + k2
1σ2
1 + k2
2σ2
2).
4.103 (a) E(Y ) = 10
R 1
0 y(1 − y)9 dy = − y(1 − y)10|1
0 +
R 1
0 (1 − y)10 dy = 1
11 .
(b) E(1 − Y ) = 1 − E(Y ) = 10
11 .
(c) V ar(Z) = V ar(1 − Y ) = V ar(Y ) = E(Y 2) − [E(Y )]2 = 10
112×12 = 0.006887.
Chapter 5
Some Discrete Probability
Distributions
5.1 This is a uniform distribution: f(x) = 1
10 , for x = 1, 2, . . . , 10.
Therefore P(X < 4) =
P3
x=1
f(x) = 3
10 .
5.2 Binomial distribution with n = 12 and p = 0.5. Hence
P(X = 3) = P(X ≤ 3) − P(X ≤ 2) = 0.0730 − 0.0193 = 0.0537.
5.3 μ =
P10
x=1
x
10 = 5.5, and σ2 =
P10
x=1
(x−5.5)2
10 = 8.25.
5.4 For n = 5 and p = 3/4, we have
(a) P(X = 2) =
5
2
(3/4)2(1/4)3 = 0.0879,
(b) P(X ≤ 3) =
P3
x=0
b(x; 5, 3/4) = 1 − P(X = 4) − P(X = 5)
= 1 −
5
4
(3/4)4(1/4)1 −
5
5
(3/4)5(1/4)0 = 0.3672.
5.5 We are considering a b(x; 20, 0.3).
(a) P(X ≥ 10) = 1 − P(X ≤ 9) = 1 − 0.9520 = 0.0480.
(b) P(X ≤ 4) = 0.2375.
(c) P(X = 5) = 0.1789. This probability is not very small so this is not a rare event.
Therefore, P = 0.30 is reasonable.
5.6 For n = 6 and p = 1/2.
(a) P(2 ≤ X ≤ 5) = P(X ≤ 5) − P(X ≤ 1) = 0.9844 − 0.1094.
(b) P(X < 3) = P(X ≤ 2) = 0.3438.
5.7 p = 0.7.
59
60 Chapter 5 Some Discrete Probability Distributions
(a) For n = 10, P(X < 5) = P(X ≤ 4) = 0.0474.
(b) For n = 20, P(X < 10) = P(X ≤ 9) = 0.0171.
5.8 For n = 8 and p = 0.6, we have
(a) P(X = 3) = b(3; 8, 0.6) = P(X ≤ 3) − P(X ≤ 2) = 0.1737 − 0.0498 = 0.1239.
(b) P(X ≥ 5) = 1 − P(X ≤ 4) = 1 − 0.4059 = 0.5941.
5.9 For n = 15 and p = 0.25, we have
(a) P(3 ≤ X ≤ 6) = P(X ≤ 6) − P(X ≤ 2) = 0.9434 − 0.2361 = 0.7073.
(b) P(X < 4) = P(X ≤ 3) = 0.4613.
(c) P(X > 5) = 1 − P(X ≤ 5) = 1 − 0.8516 = 0.1484.
5.10 From Table A.1 with n = 12 and p = 0.7, we have
(a) P(7 ≤ X ≤ 9) = P(X ≤ 9) − P(X ≤ 6) = 0.7472 − 0.1178 = 0.6294.
(b) P(X ≤ 5) = 0.0386.
(c) P(X ≥ 8) = 1 − P(X ≤ 7) = 1 − 0.2763 = 0.7237.
5.11 From Table A.1 with n = 7 and p = 0.9, we have
P(X = 5) = P(X ≤ 5) − P(X ≤ 4) = 0.1497 − 0.0257 = 0.1240.
5.12 From Table A.1 with n = 9 and p = 0.25, we have P(X < 4) = 0.8343.
5.13 From Table A.1 with n = 5 and p = 0.7, we have
P(X ≥ 3) = 1 − P(X ≤ 2) = 1 − 0.1631 = 0.8369.
5.14 (a) n = 4, P(X = 4) = 1 − 0.3439 = 0.6561.
(b) Assuming the series went to the seventh game, the probability that the Bulls won
3 of the first 6 games and then the seventh game is given by
6
3
(0.9)3(0.1)3
(0.9) = 0.0131.
(c) The probability that the Bulls win is always 0.9.
5.15 p = 0.4 and n = 5.
(a) P(X = 0) = 0.0778.
(b) P(X < 2) = P(X ≤ 1) = 0.3370.
(c) P(X > 3) = 1 − P(X ≤ 3) = 1 − 0.9130 = 0.0870.
Solutions for Exercises in Chapter 5 61
5.16 Probability of 2 or more of 4 engines operating when p = 0.6 is
P(X ≥ 2) = 1 − P(X ≤ 1) = 0.8208,
and the probability of 1 or more of 2 engines operating when p = 0.6 is
P(X ≥ 1) = 1 − P(X = 0) = 0.8400.
The 2-engine plane has a slightly higher probability for a successful flight when p = 0.6.
5.17 Since μ = np = (5)(0.7) = 3.5 and σ2 = npq = (5)(0.7)(0.3) = 1.05 with σ = 1.025.
Then μ± 2σ = 3.5± (2)(1.025) = 3.5± 2.050 or from 1.45 to 5.55. Therefore, at least
3/4 of the time when 5 people are selected at random, anywhere from 2 to 5 are of the
opinion that tranquilizers do not cure but only cover up the real problem.
5.18 (a) μ = np = (15)(0.25) = 3.75.
(b) With k = 2 and σ = √npq =
p
(15)(0.25)(0.75) = 1.677, μ ± 2σ = 3.75 ± 3.354
or from 0.396 to 7.104.
5.19 Let X1 = number of times encountered green light with P(Green) = 0.35,
X2 = number of times encountered yellow light with P(Yellow) = 0.05, and
X3 = number of times encountered red light with P(Red) = 0.60. Then
f(x1, x2, x3) =
n
x1, x2, x3
(0.35)x1(0.05)x2(0.60)x3.
5.20 (a)
10
2,5,3
(0.225)2(0.544)5(0.231)3 = 0.0749.
(b)
10
10
(0.544)10(0.456)0 = 0.0023.
(c)
10
0
(0.225)0(0.775)10 = 0.0782.
5.21 Using the multinomial distribution with required probability is
7
0, 0, 1, 4, 2
(0.02)(0.82)4(0.1)2 = 0.0095.
5.22 Using the multinomial distribution, we have
8
5,2,1
(1/2)5(1/4)2(1/4) = 21/256.
5.23 Using the multinomial distribution, we have
9
3, 3, 1, 2
(0.4)3(0.2)3(0.3)(0.2)2 = 0.0077.
5.24 p = 0.40 and n = 6, so P(X = 4) = P(X ≤ 4)−P(X ≤ 3) = 0.9590−0.8208 = 0.1382.
5.25 n = 20 and the probability of a defective is p = 0.10. So, P(X ≤ 3) = 0.8670.
62 Chapter 5 Some Discrete Probability Distributions
5.26 n = 8 and p = 0.60;
(a) P(X = 6) =
8
6
(0.6)6(0.4)2 = 0.2090.
(b) P(X = 6) = P(X ≤ 6) − P(X ≤ 5) = 0.8936 − 0.6846 = 0.2090.
5.27 n = 20 and p = 0.90;
(a) P(X = 18) = P(X ≤ 18) − P(X ≤ 17) = 0.6083 − 0.3231 = 0.2852.
(b) P(X ≥ 15) = 1 − P(X ≤ 14) = 1 − 0.0113 = 0.9887.
(c) P(X ≤ 18) = 0.6083.
5.28 n = 20;
(a) p = 0.20, P(X ≥ x) ≤ 0.5 and P(X < x) > 0.5 yields x = 4.
(b) p = 0.80, P(Y ≥ y) ≥ 0.8 and P(Y < y) < 0.2 yields y = 14.
5.29 Using the hypergeometric distribution, we get
(a)
(12
2 )(40
5 )
(52
7 )
= 0.3246.
(b) 1 −
(48
7 )
(52
7 )
= 0.4496.
5.30 P(X ≥ 1) = 1 − P(X = 0) = 1 − h(0; 15, 3, 6) = 1 −
(6
0)(9
3)
(15
3 )
= 53
65 .
5.31 Using the hypergeometric distribution, we get h(2; 9, 6, 4) =
(4
2)(5
4)
(9
6)
= 5
14 .
5.32 (a) Probability that all 4 fire = h(4; 10, 4, 7) = 1
6 .
(b) Probability that at most 2 will not fire =
P2
x=0
h(x; 10, 4, 3) = 29
30 .
5.33 h(x; 6, 3, 4) =
(4
x)( 2
3−x)
(6
3)
, for x = 1, 2, 3.
P(2 ≤ X ≤ 3) = h(2; 6, 3, 4) + h(3; 6, 3, 4) = 4
5 .
5.34 h(2; 9, 5, 4) =
(4
2)(5
3)
(9
5)
= 10
21 .
5.35 P(X ≤ 2) =
P2
x=0
h(x; 50, 5, 10) = 0.9517.
5.36 (a) P(X = 0) = h(0; 25, 3, 3) = 77
115 .
(b) P(X = 1) = h(1; 25, 3, 1) = 3
25 .
5.37 (a) P(X = 0) = b(0; 3, 3/25) = 0.6815.
Solutions for Exercises in Chapter 5 63
(b) P(1 ≤ X ≤ 3) =
P3
x=1
b(x; 3, 1/25) = 0.1153.
5.38 Since μ = (4)(3/10) = 1.2 and σ2 = (4)(3/10)(7/10)(6/9) = 504/900 with σ = 0.7483,
at least 3/4 of the time the number of defectives will fall in the interval
μ ± 2σ = 1.2 ± (2)(0.7483), or from − 0.297 to 2.697,
and at least 8/9 of the time the number of defectives will fall in the interval
μ ± 3σ = 1.2 ± (3)(0.7483) or from − 1.045 to 3.445.
5.39 Since μ = (13)(13/52) = 3.25 and σ2 = (13)(1/4)(3/4)(39/51) = 1.864 with σ = 1.365,
at least 75% of the time the number of hearts lay between
μ ± 2σ = 3.25 ± (2)(1.365) or from 0.52 to 5.98.
5.40 The binomial approximation of the hypergeometric with p = 1 − 4000/10000 = 0.6
gives a probability of
P7
x=0
b(x; 15, 0.6) = 0.2131.
5.41 Using the binomial approximation of the hypergeometric with p = 0.5, the probability
is 1 −
P2
x=0
b(x; 10, 0.5) = 0.9453.
5.42 Using the binomial approximation of the hypergeometric distribution with p = 30/150 =
0.2, the probability is 1 −
P2
x=0
b(x; 10, 0.2) = 0.3222.
5.43 Using the binomial approximation of the hypergeometric distribution with 0.7, the
probability is 1 −
P13
x=10
b(x; 18, 0.7) = 0.6077.
5.44 Using the extension of the hypergeometric distribution the probability is
13
5
13
2
13
3
13
3
52
13
= 0.0129.
5.45 (a) The extension of the hypergeometric distribution gives a probability
2
1
3
1
5
1
2
1
12
4
=
4
33
.
(b) Using the extension of the hypergeometric distribution, we have
2
1
3
1
2
2
12
4
+
2
2
3
1
2
1
12
4
+
2
1
3
2
2
1
12
4
=
8
165
.
64 Chapter 5 Some Discrete Probability Distributions
5.46 Using the extension of the hypergeometric distribution the probability is
2
2
4
1
3
2
9
5
+
2
2
4
2
3
1
9
5
+
2
2
4
3
3
0
9
5
=
17
63
.
5.47 h(5; 25, 15, 10) =
(10
5 )(15
10)
(25
15)
= 0.2315.
5.48 (a)
(2
1)(13
4 )
(15
5 )
= 0.4762.
(b)
(2
2)(13
3 )
(15
5 )
= 0.0952.
5.49 (a)
(3
0)(17
5 )
(20
5 )
= 0.3991.
(b)
(3
2)(17
3 )
(20
5 )
= 0.1316.
5.50 N = 10000, n = 30 and k = 300. Using binomial approximation to the hypergeometric
distribution with p = 300/10000 = 0.03, the probability of {X ≥ 1} can be determined
by
1 − b(0; 30, 0.03) = 1 − (0.97)30 = 0.599.
5.51 Using the negative binomial distribution, the required probability is
b∗(10; 5, 0.3) =
9
4
(0.3)5(0.7)5 = 0.0515.
5.52 From the negative binomial distribution, we obtain
b∗(8; 2, 1/6) =
7
1
(1/6)2(5/6)6 = 0.0651.
5.53 (a) P(X > 5) =
∞P x
=
6
p(x; 5) = 1 −
P5
x=0
p(x; 5) = 0.3840.
(b) P(X = 0) = p(0; 5) = 0.0067.
5.54 (a) Using the negative binomial distribution, we get
b∗(7; 3, 1/2) =
6
2
(1/2)7 = 0.1172.
(b) From the geometric distribution, we have g(4; 1/2) = (1/2)(1/2)3 = 1/16.
Solutions for Exercises in Chapter 5 65
5.55 The probability that all coins turn up the same is 1/4. Using the geometric distribution
with p = 3/4 and q = 1/4, we have
P(X < 4) =
X3
x=1
g(x; 3/4) =
X3
x=1
(3/4)(1/4)x−1 =
63
64
.
5.56 (a) Using the geometric distribution, we have g(5; 2/3) = (2/3)(1/3)4 = 2/243.
(b) Using the negative binomial distribution, we have
b∗(5; 3, 2/3) =
4
2
(2/3)3(1/3)2 =
16
81
.
5.57 Using the geometric distribution
(a) P(X = 3) = g(3; 0.7) = (0.7)(0.3)2 = 0.0630.
(b) P(X < 4) =
P3
x=1
g(x; 0.7) =
P3
x=1
(0.7)(0.3)x−1 = 0.9730.
5.58 (a) Using the Poisson distribution with x = 5 and μ = 3, we find from Table A.2 that
p(5; 3) =
X5
x=0
p(x; 3) −
X4
x=0
p(x; 3) = 0.1008.
(b) P(X < 3) = P(X ≤ 2) = 0.4232.
(c) P(X ≥ 2) = 1 − P(X ≤ 1) = 0.8009.
5.59 (a) P(X ≥ 4) = 1 − P(X ≤ 3) = 0.1429.
(b) P(X = 0) = p(0; 2) = 0.1353.
5.60 (a) P(X < 4) = P(X ≤ 3) = 0.1512.
(b) P(6 ≤ X ≤ 8) = P(X ≤ 8) − P(X ≤ 5) = 0.4015.
5.61 (a) Using the negative binomial distribution, we obtain
b∗(6; 4, 0.8) =
5
3
(0.8)4(0.2)2 = 0.1638.
(b) From the geometric distribution, we have g(3; 0.8) = (0.8)(0.2)2 = 0.032.
5.62 (a) Using the Poisson distribution with μ = 12, we find from Table A.2 that
P(X < 7) = P(X ≤ 6) = 0.0458.
(b) Using the binomial distribution with p = 0.0458, we get
b(2; 3, 0.0458) =
3
2
(0.0458)2(0.9542) = 0.0060.
66 Chapter 5 Some Discrete Probability Distributions
5.63 (a) Using the Poisson distribution with μ = 5, we find
P(X > 5) = 1 − P(X ≤ 5) = 1 − 0.6160 = 0.3840.
(b) Using the binomial distribution with p = 0.3840, we get
b(3; 4, 0.384) =
4
3
(0.3840)3(0.6160) = 0.1395.
(c) Using the geometric distribution with p = 0.3840, we have
g(5; 0.384) = (0.394)(0.616)4 = 0.0553.
5.64 μ = np = (2000)(0.002) = 4, so P(X < 5) = P(X ≤ 4) ≈
P4
x=0
p(x; 4) = 0.6288.
5.65 μ = np = (10000)(0.001) = 10, so
P(6 ≤ X ≤ 8) = P(X ≤ 8) − P(X ≤ 5) ≈
X8
x=0
p(x; 10) −
X5
x=0
p(x; 10) = 0.2657.
5.66 (a) μ = np = (1875)(0.004) = 7.5, so P(X < 5) = P(X ≤ 4) ≈ 0.1321.
(b) P(8 ≤ X ≤ 10) = P(X ≤ 10) − P(X ≤ 7) ≈ 0.8622 − 0.5246 = 0.3376.
5.67 (a) μ = (2000)(0.002) = 4 and σ2 = 4.
(b) For k = 2, we have μ ± 2σ = 4 ± 4 or from 0 to 8.
5.68 (a) μ = (10000)(0.001) = 10 and σ2 = 10.
(b) For k = 3, we have μ ± 3σ = 10 ± 3√10 or from 0.51 to 19.49.
5.69 (a) P(X ≤ 3|λt = 5) = 0.2650.
(b) P(X > 1|λt = 5) = 1 − 0.0404 = 0.9596.
5.70 (a) P(X = 4|λt = 6) = 0.2851 − 0.1512 = 0.1339.
(b) P(X ≥ 4|λt = 6) = 1 − 0.1512 = 0.8488.
(c) P(X ≥ 75|λt = 72) = 1 −
P74
x=0
p(x; 74) = 0.3773.
5.71 (a) P(X > 10|λt = 14) = 1 − 0.1757 = 0.8243.
(b) λt = 14.
5.72 μ = np = (1875)(0.004) = 7.5.
5.73 μ = (4000)(0.001) = 4.
Solutions for Exercises in Chapter 5 67
5.74 μ = 1 and σ2 = 0.99.
5.75 μ = λt = (1.5)(5) = 7.5 and P(X = 0|λt = 7.5) = e−7.5 = 5.53 × 10−4.
5.76 (a) P(X ≤ 1|λt = 2) = 0.4060.
(b) μ = λt = (2)(5) = 10 and P(X ≤ 4|λt = 10) = 0.0293.
5.77 (a) P(X > 10|λt = 5) = 1 − P(X ≤ 10|λt = 5) = 1 − 0.9863 = 0.0137.
(b) μ = λt = (5)(3) = 15, so P(X > 20|λt = 15) = 1 − P(X ≤ 20|λ = 15) =
1 − 0.9170 = 0.0830.
5.78 p = 0.03 with a g(x; 0.03). So, P(X = 16) = (0.03)(1 − 0.03)15 = 0.0190 and
μ = 1
0.03 − 1 = 32.33.
5.79 So, Let Y = number of shifts until it fails. Then Y follows a geometric distribution
with p = 0.10. So,
P(Y ≤ 6) = g(1; 0.1) + g(2; 0.1) + · · · + g(6; 0.1)
= (0.1)[1 + (0.9) + (0.9)2 + · · · + (0.9)5] = 0.4686.
5.80 (a) The number of people interviewed before the first refusal follows a geometric
distribution with p = 0.2. So.
P(X ≥ 51) =
∞X
x=51
(0.2)(1 − 0.2)x = (0.2)
(1 − 0.2)50
1 − (1 − 0.2)
= 0.00001,
which is a very rare event.
(b) μ = 1
0.2 − 1 = 4.
5.81 n = 15 and p = 0.05.
(a) P(X ≥ 2) = 1 − P(X ≤ 1) = 1 −
P1
x=0
15
x
(0.05)x(1 − 0.05)15−x = 1 − 0.8290 =
0.1710.
(b) p = 0.07. So, P(X ≤ 1) =
P1
x=0
15
x
(0.07)x(1 − 0.07)15−x = 1 − 0.7168 = 0.2832.
5.82 n = 100 and p = 0.01.
(a) P(X > 3) = 1 − P(X ≤ 3) = 1 −
P3
x=0
100
x
(0.01)x(1 − 0.01)100−x = 1 − 0.9816 =
0.0184.
(b) For p = 0.05, P(X ≤ 3) =
P3
x=0
100
x
(0.05)x(1 − 0.05)100−x = 0.2578.
68 Chapter 5 Some Discrete Probability Distributions
5.83 Using the extension of the hypergeometric distribution, the probability is
5
2
7
3
4
1
3
1
4
2
5
2
= 0.0308.
5.84 λ = 2.7 call/min.
(a) P(X ≤ 4) =
P4
x=0
e−2.7(2.7)x
x! = 0.8629.
(b) P(X ≤ 1) =
P1
x=0
e−2.7(2.7)x
x! = 0.2487.
(c) λt = 13.5. So,
P(X > 10) = 1 − P(X ≤ 10) = 1 −
P10
x=0
e−13.5(13.5)x
x! = 1 − 0.2971 = 0.7129.
5.85 n = 15 and p = 0.05.
(a) P(X = 5) =
15
5
(0.05)5(1 − 0.05)10 = 0.000562.
(b) I would not believe the claim of 5% defective.
5.86 λ = 0.2, so λt = (0.2)(5) = 1.
(a) P(X ≤ 1) =
P1
x=0
e−1(1)x
x! = 2e−1 = 0.7358. Hence, P(X > 1) = 1−0.7358 = 0.2642.
(b) λ = 0.25, so λt = 1.25. P(X|le1) =
P1
x=0
e−1.25(1.25)x
x! = 0.6446.
5.87 (a) 1 − P(X ≤ 1) = 1 −
P1
x=0
100
x
(0.01)x(0.99)100−x = 1 − 0.7358 = 0.2642.
(b) P(X ≤ 1) =
P1
x=0
100
x
(0.05)x(0.95)100−x = 0.0371.
5.88 (a) 100 visits/60 minutes with λt = 5 visits/3 minutes. P(X = 0) = e−550
0! = 0.0067.
(b) P(X > 5) = 1 −
P5
x=0
e55x
x! = 1 − 0.6160 = 0.3840.
5.89 (a) P(X ≥ 1) = 1 − P(X = 0) = 1 −
4
0
1
6
0 5
6
4
= 1 − 0.4822 = 0.5177.
(b) P(X ≥ 1) = 1 − P(X = 0) = 1 −
24
0
1
36
0
35
36
24
= 1 − 0.5086 = 0.4914.
5.90 n = 5 and p = 0.4; P(X ≥ 3) = 1 − P(X ≤ 2) = 1 − 0.6826 = 0.3174.
5.91 (a) μ = bp = (200)(0.03) = 6.
(b) σ2 = npq = 5.82.
Solutions for Exercises in Chapter 5 69
(c) P(X = 0) = e6(6)0
0! = 0.0025 (using the Poisson approximation).
P(X = 0) = (0.97)200 = 0.0023 (using the binomial distribution).
5.92 (a) p10q0 = (0.99)10 = 0.9044.
(b) p10q12−10 = (0.99)10(0.01)2 = (0.9044)(0.0001) = 0.00009.
5.93 n = 75 with p = 0.999.
(a) X = the number of trials, and P(X = 75) = (0.999)75(0.001)0 = 0.9277.
(b) Y = the number of trials before the first failure (geometric distribution), and
P(Y = 20) = (0.001)(0.999)19 = 0.000981.
(c) 1 − P(no failures) = 1 − (0.001)0(0.999)10 = 0.01.
5.94 (a)
10
1
pq9 = (10)(0.25)(0.75)9 = 0.1877.
(b) Let X be the number of drills until the first success. X follows a geometric
distribution with p = 0.25. So, the probability of having the first 10 drills being
failure is q10 = (0.75)10 = 0.056. So, there is a small prospects for bankruptcy.
Also, the probability that the first success appears in the 11th drill is pq10 = 0.014
which is even smaller.
5.95 It is a negative binomial distribution.
x−1
k−1
pkqx−k =
6−1
2−1
(0.25)2(0.75)4 = 0.0989.
5.96 It is a negative binomial distribution.
x−1
k−1
pkqx−k =
4−1
2−1
(0.5)2(0.5)2 = 0.1875.
5.97 n = 1000 and p = 0.01, with μ = (1000)(0.01) = 10. P(X < 7) = P(X ≤ 6) = 0.1301.
5.98 n = 500;
(a) If p = 0.01,
P(X ≥ 15) = 1 − P(X ≤ 14) = 1 −
X14
x=0
500
x
(0.01)x(0.99)500−x = 0.00021.
This is a very rare probability and thus the original claim that p = 0.01 is questionable.
(b) P(X = 3) =
500
3
(0.01)3(0.99)497 = 0.1402.
(c) For (a), if p = 0.01, μ = (500)(0.01) = 5. So
P(X ≥ 15) = 1 − P(X ≤ 14) = 1 − 0.9998 = 0.0002.
For (b),
P(X = 3) = 0.2650 − 0.1247 = 0.1403.
5.99 N = 50 and n = 10.
70 Chapter 5 Some Discrete Probability Distributions
(a) k = 2; P(X ≥ 1) = 1 − P(X = 0) = 1 −
(2
0)(48
10)
(50
10)
= 1 − 0.6367 = 0.3633.
(b) Even though the lot contains 2 defectives, the probability of reject the lot is not
very high. Perhaps more items should be sampled.
(c) μ = (10)(2/50) = 0.4.
5.100 Suppose n items need to be sampled. P(X ≥ 1) = 1−
(2
0)(48
n )
(50
n )
= 1 − (50−n)(49−n)
(50)(49) ≥ 0.9.
The solution is n = 34.
5.101 Define X = number of screens will detect. Then X ∼ b(x; 3, 0.8).
(a) P(X = 0) = (1 − 0.8)3 = 0.008.
(b) P(X = 1) = (3)(0.2)2(0.8) = 0.096.
(c) P(X ≥ 2) = P(X = 2) + P(X = 3) = (3)(0.8)2(0.2) + (0.8)3 = 0.896.
5.102 (a) P(X = 0) = (1 − 0.8)n ≤ 0.0001 implies that n ≥ 6.
(b) (1 − p)3 ≤ 0.0001 implies p ≥ 0.9536.
5.103 n = 10 and p = 2
50 = 0.04.
P(X ≥ 1) = 1 − P(X = 0) ≈ 1 −
10
0
(0.04)0(1 − 0.04)10 = 1 − 0.6648 = 0.3351.
The approximation is not that good due to n
N = 0.2 is too large.
5.104 (a) P =
(2
2)(3
0)
(5
2)
= 0.1.
(b) P =
(2
1)(1
1)
(5
2)
= 0.2.
5.105 n = 200 with p = 0.00001.
(a) P(X ≥ 5) = 1 − P(X ≤ 4) = 1 −
P4
x=0
200
x
(0.00001)x(1 − 0.00001)200−x ≈ 0. This
is a rare event. Therefore, the claim does not seem right.
(b) μ = np = (200)(0.00001) = 0.02. Using Poisson approximation,
P(X ≥ 5) = 1 − P(X ≤ 4) ≈ 1 −
X4
x=0
e−0.02 (0.02)x
x!
= 0.
Chapter 6
Some Continuous Probability
Distributions
6.1 (a) Area=0.9236.
(b) Area=1 − 0.1867 = 0.8133.
(c) Area=0.2578 − 0.0154 = 0.2424.
(d) Area=0.0823.
(e) Area=1 − 0.9750 = 0.0250.
(f) Area=0.9591 − 0.3156 = 0.6435.
6.2 (a) The area to the left of z is 1−0.3622 = 0.6378 which is closer to the tabled value
0.6368 than to 0.6406. Therefore, we choose z = 0.35.
(b) From Table A.3, z = −1.21.
(c) The total area to the left of z is 0.5000+0.4838=0.9838. Therefore, from Table
A.3, z = 2.14.
(d) The distribution contains an area of 0.025 to the left of −z and therefore a total
area of 0.025+0.95=0.975 to the left of z. From Table A.3, z = 1.96.
6.3 (a) From Table A.3, k = −1.72.
(b) Since P(Z > k) = 0.2946, then P(Z < k) = 0.7054/ From Table A.3, we find
k = 0.54.
(c) The area to the left of z = −0.93 is found from Table A.3 to be 0.1762. Therefore,
the total area to the left of k is 0.1762+0.7235=0.8997, and hence k = 1.28.
6.4 (a) z = (17 − 30)/6 = −2.17. Area=1 − 0.0150 = 0.9850.
(b) z = (22 − 30)/6 = −1.33. Area=0.0918.
(c) z1 = (32−3)/6 = 0.33, z2 = (41−30)/6 = 1.83. Area = 0.9664−0.6293 = 0.3371.
(d) z = 0.84. Therefore, x = 30 + (6)(0.84) = 35.04.
71
72 Chapter 6 Some Continuous Probability Distributions
(e) z1 = −1.15, z2 = 1.15. Therefore, x1 = 30 + (6)(−1.15) = 23.1 and x2 =
30 + (6)(1.15) = 36.9.
6.5 (a) z = (15 − 18)/2.5 = −1.2; P(X < 15) = P(Z < −1.2) = 0.1151.
(b) z = −0.76, k = (2.5)(−0.76) + 18 = 16.1.
(c) z = 0.91, k = (2.5)(0.91) + 18 = 20.275.
(d) z1 = (17 − 18)/2.5 = −0.4, z2 = (21 − 18)/2.5 = 1.2;
P(17 < X < 21) = P(−0.4 < Z < 1.2) = 0.8849 − 0.3446 = 0.5403.
6.6 z1 = [(μ − 3σ) − μ]/σ = −3, z2 = [(μ + 3σ) − μ]/σ = 3;
P(μ − 3σ < Z < μ + 3σ) = P(−3 < Z < 3) = 0.9987 − 0.0013 = 0.9974.
6.7 (a) z = (32 − 40)/6.3 = −1.27; P(X > 32) = P(Z > −1.27) = 1 − 0.1020 = 0.8980.
(b) z = (28 − 40)/6.3 = −1.90, P(X < 28) = P(Z < −1.90) = 0.0287.
(c) z1 = (37 − 40)/6.3 = −0.48, z2 = (49 − 40)/6.3 = 1.43;
So, P(37 < X < 49) = P(−0.48 < Z < 1.43) = 0.9236 − 0.3156 = 0.6080.
6.8 (a) z = (31.7 − 30)/2 = 0.85; P(X > 31.7) = P(Z > 0.85) = 0.1977.
Therefore, 19.77% of the loaves are longer than 31.7 centimeters.
(b) z1 = (29.3 − 30)/2 = −0.35, z2 = (33.5 − 30)/2 = 1.75;
P(29.3 < X < 33.5) = P(−0.35 < Z < 1.75) = 0.9599 − 0.3632 = 0.5967.
Therefore, 59.67% of the loaves are between 29.3 and 33.5 centimeters in length.
(c) z = (25.5 − 30)/2 = −2.25; P(X < 25.5) = P(Z < −2.25) = 0.0122.
Therefore, 1.22% of the loaves are shorter than 25.5 centimeters in length.
6.9 (a) z = (224 − 200)/15 = 1.6. Fraction of the cups containing more than 224 millimeters
is P(Z > 1.6) = 0.0548.
(b) z1 = (191 − 200)/15 = −0.6, Z2 = (209 − 200)/15 = 0.6;
P(191 < X < 209) = P(−0.6 < Z < 0.6) = 0.7257 − 0.2743 = 0.4514.
(c) z = (230 − 200)/15 = 2.0; P(X > 230) = P(Z > 2.0) = 0.0228. Therefore,
(1000)(0.0228) = 22.8 or approximately 23 cups will overflow.
(d) z = −0.67, x = (15)(−0.67) + 200 = 189.95 millimeters.
6.10 (a) z = (10.075 − 10.000)/0.03 = 2.5; P(X > 10.075) = P(Z > 2.5) = 0.0062.
Therefore, 0.62% of the rings have inside diameters exceeding 10.075 cm.
(b) z1 = (9.97 − 10)/0.03 = −1.0, z2 = (10.03 − 10)/0.03 = 1.0;
P(9.97 < X < 10.03) = P(−1.0 < Z < 1.0) = 0.8413 − 0.1587 = 0.6826.
(c) z = −1.04, x = 10 + (0.03)(−1.04) = 9.969 cm.
6.11 (a) z = (30 − 24)/3.8 = 1.58; P(X > 30) = P(Z > 1.58) = 0.0571.
(b) z = (15 − 24)/3.8 = −2.37; P(X > 15) = P(Z > −2.37) = 0.9911. He is late
99.11% of the time.
Solutions for Exercises in Chapter 6 73
(c) z = (25 − 24)/3.8 = 0.26; P(X > 25) = P(Z > 0.26) = 0.3974.
(d) z = 1.04, x = (3.8)(1.04) + 24 = 27.952 minutes.
(e) Using the binomial distribution with p = 0.0571, we get
b(2; 3, 0.0571) =
3
2
(0.0571)2(0.9429) = 0.0092.
6.12 μ = 99.61 and σ = 0.08.
(a) P(99.5 < X < 99.7) = P(−1.375 < Z < 1.125) = 0.8697 − 0.08455 = 0.7852.
(b) P(Z > 1.645) = 0.05; x = (1.645)(0.08) + 99.61 = 99.74.
6.13 z = −1.88, x = (2)(−1.88) + 10 = 6.24 years.
6.14 (a) z = (159.75 − 174.5)/6.9 = −2.14; P(X < 159.75) = P(Z < −2.14) = 0.0162.
Therefore, (1000)(0.0162) = 16 students.
(b) z1 = (171.25 − 174.5)/6.9 = −0.47, z2 = (182.25 − 174.5)/6.9 = 1.12.
P(171.25 < X < 182.25) = P(−0.47 < Z < 1.12) = 0.8686 − 0.3192 = 0.5494.
Therefore, (1000)(0.5494) = 549 students.
(c) z1 = (174.75 − 174.5)/6.9 = 0.04, z2 = (175.25 − 174.5)/6.9 = 0.11.
P(174.75 < X < 175.25) = P(0.04 < Z < 0.11) = 0.5438 − 0.5160 = 0.0278.
Therefore, (1000)(0.0278)=28 students.
(d) z = (187.75 − 174.5)/6.9 = 1.92; P(X > 187.75) = P(Z > 1.92) = 0.0274.
Therefore, (1000)(0.0274) = 27 students.
6.15 μ = $15.90 and σ = $1.50.
(a) 51%, since P(13.75 < X < 16.22) = P
13.745−15.9
1.5 < Z < 16.225−15.9
1.5
= P(−1.437 < Z < 0.217) = 0.5871 − 0.0749 = 0.5122.
(b) $18.36, since P(Z > 1.645) = 0.05; x = (1.645)(1.50) + 15.90 + 0.005 = 18.37.
6.16 (a) z = (9.55 − 8)/0.9 = 1.72. Fraction of poodles weighing over 9.5 kilograms =
P(X > 9.55) = P(Z > 1.72) = 0.0427.
(b) z = (8.65 − 8)/0.9 = 0.72. Fraction of poodles weighing at most 8.6 kilograms =
P(X < 8.65) = P(Z < 0.72) = 0.7642.
(c) z1 = (7.25 − 8)/0.9 = −0.83 and z2 = (9.15 − 8)/0.9 = 1.28.
Fraction of poodles weighing between 7.3 and 9.1 kilograms inclusive
= P(7.25 < X < 9.15) = P(−0.83 < Z < 1.28) = 0.8997 − 0.2033 = 0.6964.
6.17 (a) z = (10, 175 − 10, 000)/100 = 1.75. Proportion of components exceeding 10.150
kilograms in tensile strength= P(X > 10, 175) = P(Z > 1.75) = 0.0401.
(b) z1 = (9, 775 − 10, 000)/100 = −2.25 and z2 = (10, 225 − 10, 000)/100 = 2.25.
Proportion of components scrapped= P(X < 9, 775) + P(X > 10, 225) = P(Z <
−2.25) + P(Z > 2.25) = 2P(Z < −2.25) = 0.0244.
74 Chapter 6 Some Continuous Probability Distributions
6.18 (a) x1 = μ + 1.3σ and x2 = μ − 1.3σ. Then z1 = 1.3 and z2 = −1.3. P(X >
μ+1.3σ)+P(X < 1.3σ) = P(Z > 1.3)+P(Z < −1.3) = 2P(Z < −1.3) = 0.1936.
Therefore, 19.36%.
(b) x1 = μ+0.52σ and x2 = μ−0.52σ. Then z1 = 0.52 and z2 = −0.52. P(μ−0.52σ <
X < μ + 0.52σ) = P(−0.52 < Z < 0.52) = 0.6985 − 0.3015 = 0.3970. Therefore,
39.70%.
6.19 z = (94.5 − 115)/12 = −1.71; P(X < 94.5) = P(Z < −1.71) = 0.0436. Therefore,
(0.0436)(600) = 26 students will be rejected.
6.20 f(x) = 1
B−A for A ≤ x ≤ B.
(a) μ =
R B
A
x
B−A dx = B2−A2
2(B−A) = A+B
2 .
(b) E(X2) =
R B
A
x2
B−A dx = B3−A3
3(B−A) .
So, σ2 = B3−A3
3(B−A) −
A+B
2
2
= 4(B2+AB+A2)−3(B2+2AB+A2)
12 = B2−2AB+A2
12 = (B−A)2
12 .
6.21 A = 7 and B = 10.
(a) P(X ≤ 8.8) = 8.8−7
3 = 0.60.
(b) P(7.4 < X < 9.5) = 9.5−7.4
3 = 0.70.
(c) P(X ≥ 8.5) = 10−8.5
3 = 0.50.
6.22 (a) P(X > 7) = 10−7
10 = 0.3.
(b) P(2 < X < 7) = 7−2
10 = 0.5.
6.23 (a) From Table A.1 with n = 15 and p = 0.2 we have
P(1 ≤ X ≤ 4) =
P4
x=0
b(x; 15, 0.2) − b(0; 15, 0.2) = 0.8358 − 0.0352 = 0.8006.
(b) By the normal-curve approximation we first find
μ = np = 3 and then σ2 = npq = (15)(0.2)(0.8) = 2.4. Then σ = 1.549.
Now, z1 = (0.5 − 3)/1.549 = −1.61 and z2 = (4.5 − 3)/1.549 = 0.97.
Therefore, P(1 ≤ X ≤ 4) = P(−1.61 ≤ Z ≤ 0.97) = 0.8340 − 0.0537 = 0.7803.
6.24 μ = np = (400)(1/2) = 200, σ = √npq =
p
(400)(1/2)(1/2) = 10.
(a) z1 = (184.5 − 200)/10 = −1.55 and z2 = (210.5 − 200)/10 = 1.05.
P(184.5 < X < 210.5) = P(−1.55 < Z < 1.05) = 0.8531 − 0.0606 = 0.7925.
(b) z1 = (204.5 − 200)/10 = 0.45 and z2 = (205.5 − 200)/10 = 0.55.
P(204.5 < X < 205.5) = P(0.45 < Z < 0.55) = 0.7088 − 0.6736 = 0.0352.
(c) z1 = (175.5 − 200)/10 = −2.45 and z2 = (227.5 − 200)/10 = 2.75.
P(X < 175.5) + P(X > 227.5) = P(Z < −2.45) + P(Z > 2.75)
= P(Z < −2.45) + 1 − P(Z < 2.75) = 0.0071 + 1 − 0.9970 = 0.0101.
Solutions for Exercises in Chapter 6 75
6.25 n = 100.
(a) p = 0.01 with μ = (100)(0.01) = 1 and σ =
p
(100)(0.01)(0.99) = 0.995.
So, z = (0.5 − 1)/0.995 = −0.503. P(X ≤ 0) ≈ P(Z ≤ −0.503) = 0.3085.
(b) p = 0.05 with μ = (100)(0.05) = 5 and σ =
p
(100)(0.05)(0.95) = 2.1794.
So, z = (0.5 − 5)/2.1794 = −2.06. P(X ≤ 0) ≈ P(X ≤ −2.06) = 0.0197.
6.26 μ = np = (100)(0.1) = 10 and σ =
p
(100)(0.1)(0.9) = 3.
(a) z = (13.5 − 10)/3 = 1.17; P(X > 13.5) = P(Z > 1.17) = 0.1210.
(b) z = (7.5 − 10)/3 = −0.83; P(X < 7.5) = P(Z < −0.83) = 0.2033.
6.27 μ = (100)(0.9) = 90 and σ =
p
(100)(0.9)(0.1) = 3.
(a) z1 = (83.5 − 90)/3 = −2.17 and z2 = (95.5 − 90)/3 = 1.83.
P(83.5 < X < 95.5) = P(−2.17 < Z < 1.83) = 0.9664 − 0.0150 = 0.9514.
(b) z = (85.5 − 90)/3 = −1.50; P(X < 85.5) = P(Z < −1.50) = 0.0668.
6.28 μ = (80)(3/4) = 60 and σ =
p
(80)(3/4)(1/4) = 3.873.
(a) z = (49.5 − 60)/3.873 = −2.71; P(X > 49.5) = P(Z > −2.71) = 1 − 0.0034 =
0.9966.
(b) z = (56.5 − 60)/3.873 = −0.90; P(X < 56.5) = P(Z < −0.90) = 0.1841.
6.29 μ = (1000)(0.2) = 200 and σ =
p
(1000)(0.2)(0.8) = 12.649.
(a) z1 = (169.5 − 200)/12.649 = −2.41 and z2 = (185.5 − 200)/12.649 = −1.15.
P(169.5 < X < 185.5) = P(−2.41 < Z < −1.15) = 0.1251 − 0.0080 = 0.1171.
(b) z1 = (209.5 − 200)/12.649 = 0.75 and z2 = (225.5 − 200)/12.649 = 2.02.
P(209.5 < X < 225.5) = P(0.75 < Z < 2.02) = 0.9783 − 0.7734 = 0.2049.
6.30 (a) μ = (100)(0.8) = 80 and σ =
p
(100)(0.8)(0.2) = 4 with z = (74.5 − 80)/4 =
−1.38.
P(Claim is rejected when p = 0.8) = P(Z < −1.38) = 0.0838.
(b) μ = (100)(0.7) = 70 and σ =
p
(100)(0.7)(0.3) = 4.583 with z = (74.5 −
70)/4.583 = 0.98.
P(Claim is accepted when p = 0.7) = P(Z > 0.98) = 1 − 0.8365 = 0.1635.
6.31 μ = (180)(1/6) = 30 and σ =
p
(180)(1/6)(5/6) = 5 with z = (35.5 − 30)/5 = 1.1.
P(X > 35.5) = P(Z > 1.1) = 1 − 0.8643 = 0.1357.
6.32 μ = (200)(0.05) = 10 and σ =
p
(200)(0.05)(0.95) = 3.082 with
z = (9.5 − 10)/3.082 = −0.16. P(X < 10) = P(Z < −0.16) = 0.4364.
6.33 μ = (400)(1/10) = 40 and σ =
p
(400)(1/10)(9/10) = 6.
76 Chapter 6 Some Continuous Probability Distributions
(a) z = (31.5 − 40)/6 = −1.42; P(X < 31.5) = P(Z < −1.42) = 0.0778.
(b) z = (49.5 − 40)/6 = 1.58; P(X > 49.5) = P(Z > 1.58) = 1 − 0.9429 = 0.0571.
(c) z1 = (34.5 − 40)/6 = −0.92 and z2 = (46.5 − 40)/6 = 1.08;
P(34.5 < X < 46.5) = P(−0.92 < Z < 1.08) = 0.8599 − 0.1788 = 0.6811.
6.34 μ = (180)(1/6) = 30 and σ =
p
(180)(1/6)(5/6) = 5.
(a) z = (24.5 − 30)/5 = −1.1; P(X > 24.5) = P(Z > −1.1) = 1 − 0.1357 = 0.8643.
(b) z1 = (32.5 − 30)/5 = 0.5 and z2 = (41.5 − 30)/5 = 2.3.
P(32.5 < X < 41.5) = P(0.5 < Z < 2.3) = 0.9893 − 0.6915 = 0.2978.
(c) z1 = (29.5 − 30)/5 = −0.1 and z2 = (30.5 − 30)/5 = 0.1.
P(29.5 < X < 30.5) = P(−0.1 < Z < 0.1) = 0.5398 − 0.4602 = 0.0796.
6.35 (a) p = 0.05, n = 100 with μ = 5 and σ =
p
(100)(0.05)(0.95) = 2.1794.
So, z = (2.5 − 5)/2.1794 = −1.147; P(X ≥ 2) ≈ P(Z ≥ −1.147) = 0.8749.
(b) z = (10.5 − 5)/2.1794 = 2.524; P(X ≥ 10) ≈ P(Z > 2.52) = 0.0059.
6.36 n = 200; X = The number of no shows with p = 0.02. z = 3−0.5−4 √(200)(0.02)(0.98)
= −0.76.
Therefore, P(airline overbooks the flight) = 1 − P(X ≥ 3) ≈ 1 − P(Z > −0.76) =
0.2236.
6.37 (a) P(X ≥ 230) = P
Z > 230−170
30
= 0.0228.
(b) Denote by Y the number of students whose serum cholesterol level exceed 230
among the 300. Then Y ∼ b(y; 300, 0.0228 with μ = (300)(0.0228) = 6.84 and
σ =
p
(300)(0.0228)(1 − 0.0228) = 2.5854. So, z = 8−0.5−6.84
2.5854 = 0.26 and
P(X ≥ 8) ≈ P(Z > 0.26) = 0.3974.
6.38 (a) Denote by X the number of failures among the 20. X ∼ b(x; 20, 0.01) and P(X >
1) = 1−b(0; 20, 0.01)−b(1; 20, 0.01) = 1−
20
0
(0.01)0(0.99)20−
20
1
(0.01)(0.99)19 =
0.01686.
(b) n = 500 and p = 0.01 with μ = (500)(0.01) = 5 and σ =
p
(500)(0.01)(0.99) =
2.2249. So, P(more than 8 failures) ≈ P(Z > (8.5−5)/2.2249) = P(Z > 1.57) =
1 − 0.9418 = 0.0582.
6.39 P(1.8 < X < 2.4) =
R 2.4
1.8 xe−x dx = [−xe−x − e−x]|2.4
1.8 = 2.8e−1.8 − 3.4e−2.4 = 0.1545.
6.40 P(X > 9) = 1
9
R
∞
9 x−x/3 dx =
−x
3 e−x/3 − e−x/3
∞9
= 4e−3 = 0.1992.
6.41 Setting α = 1/2 in the gamma distribution and integrating, we have
1
√β(1/2)
Z
∞
0
x−1/2e−x/ dx = 1.
Solutions for Exercises in Chapter 6 77
Substitute x = y2/2, dx = y dy, to give
(1/2) =
√2
√β
Z
∞
0
e−y2/2 dy = 2√π
1
√2π√β
Z
∞
0
e−y2/2 dy
= √π,
since the quantity in parentheses represents one-half of the area under the normal curve
n(y; 0,√β).
6.42 (a) P(X < 1) = 4
R 1
0 xe−2x dx = [−2xe−2x − e−2x]|1
0 = 1 − 3e−2 = 0.5940.
(b) P(X > 2) = 4
R
∞
0 xe−2x dx = [−2xe−2x − e−2x]|∞2 = 5e−4 = 0.0916.
6.43 (a) μ = αβ = (2)(3) = 6 million liters; σ2 = αβ2 = (2)(9) = 18.
(b) Water consumption on any given day has a probability of at least 3/4 of falling
in the interval μ ± 2σ = 6 ± 2√18 or from −2.485 to 14.485. That is from 0 to
14.485 million liters.
6.44 (a) μ = αβ = 6 and σ2 = αβ2 = 12. Substituting α = 6/β into the variance formula
we find 6β = 12 or β = 2 and then α = 3.
(b) P(X > 12) = 1
16
R
∞
12 x2e−x/2 dx. Integrating by parts twice gives
P(X > 12) =
1
16
−2x2e−x/2 − 8xe−x/2 − 16e−x/2 ∞
12 = 25e−6 = 0.0620.
6.45 P(X < 3) = 1
4
R 3
0 e−x/4 dx = −e−x/4
3
0 = 1 − e−3/4 = 0.5276.
Let Y be the number of days a person is served in less than 3 minutes. Then
P(Y ≥ 4) =
P6
x=4
b(y; 6, 1 − e−3/4) =
6
4
(0.5276)4(0.4724)2 +
6
5
(0.5276)5(0.4724)
+
6
6
(0.5276)6 = 0.3968.
6.46 P(X < 1) = 1
2
R 1
0 e−x/2 dx = −e−x/2
1
0 = 1 − e−1/2 = 0.3935. Let Y be the number
of switches that fail during the first year. Using the normal approximation we find
μ = (100)(0.3935) = 39.35, σ =
p
(100)(0.3935)(0.6065) = 4.885, and z = (30.5 − 39.35)/4.885 = −1.81. Therefore, P(Y ≤ 30) = P(Z < −1.81) = 0.0352.
6.47 (a) E(X) =
R
∞
0 x2e−x2/2 dx = −xe−x2/2
∞
0
+
R
∞
0 e−x2/2 dx
= 0 + √2π · 1 √2
R
∞
0 e−x2/2 dx =
√2
2 =
p
2 = 1.2533.
(b) P(X > 2) =
R
∞
2 xe−x2/2 dx = −e−x2/2
∞
2
= e−2 = 0.1353.
6.48 μ = E(T) = αβ
R
∞
0 t e− t dt. Let y = αt , then dy = αβt −1 dt and t = (y/α)1/ .
Then
μ =
Z
∞
0
(y/α)1/ e−y dy = α−1/
Z
∞
0
y(1+1/ )−1e−y dy = α−1/ (1 + 1/β).
78 Chapter 6 Some Continuous Probability Distributions
E(T2) = αβ
Z
∞
0
t +1e− t
dt =
Z
∞
0
(y/α)2/ e−y dy = α−2/
Z
∞
0
y(1+2/ )−1e−y dy
= α−2/ (1 + 2/β).
So, σ2 = E(T2) − μ2 = α−2/ {(1 + 2/β) − [(1 + 1/β)]2}.
6.49 R(t) = ce−R 1/√t dt = ce−2√t. However, R(0) = 1 and hence c = 1. Now
f(t) = Z(t)R(t) = e−2√t/√t, t > 0,
and
P(T > 4) =
Z
∞
4
e−2√t/√t dt = −e−2√t
∞
4
= e−4 = 0.0183.
6.50 f(x) = 12x2(1 − x), 0 < x < 1. Therefore,
P(X > 0.8) = 12
Z 1
0.8
x2(1 − x) dx = 0.1808.
6.51 α = 5; β = 10;
(a) αβ = 50.
(b) σ2 = αβ2 = 500; so σ = √500 = 22.36.
(c) P(X > 30) = 1
( )
R
∞
30 x −1e−x/ dx. Using the incomplete gamma with y =
x/β, then
1 − P(X ≤ 30) = 1 − P(Y ≤ 3) = 1 −
Z 3
0
y4e−y
(5)
dy = 1 − 0.185 = 0.815.
6.52 αβ = 10; σ =
p
αβ2√50 = 7.07.
(a) Using integration by parts,
P(X ≤ 50) =
1
β (α)
Z 50
0
x −1e−x/ dx =
1
25
Z 50
0
xe−x/5 dx = 0.9995.
(b) P(X < 10) = 1
( )
R 10
0 x −1e−x/ dx. Using the incomplete gamma with y =
x/β, we have
P(X < 10) = P(Y < 2) =
Z 2
0
ye−y dy = 0.5940.
6.53 μ = 3 seconds with f(x) = 1
3e−x/3 for x > 0.
Solutions for Exercises in Chapter 6 79
(a) P(X > 5) =
R
∞
5
1
3e−x/3 dx = 1
3
−3e−x/3
∞5
= e−5/3 = 0.1889.
(b) P(X > 10) = e−10/3 = 0.0357.
6.54 P(X > 270) = 1 −
ln 270−4
2
= 1 − (0.7992) = 0.2119.
6.55 μ = E(X) = e4+4/2 = e6; σ2 = e8+4(e4 − 1) = e12(e4 − 1).
6.56 β = 1/5 and α = 10.
(a) P(X > 10) = 1 − P(X ≤ 10) = 1 − 0.9863 = 0.0137.
(b) P(X > 2) before 10 cars arrive.
P(X ≤ 2) =
Z 2
0
1
β
x −1e−x/
(α)
dx.
Given y = x/β, then
P(X ≤ 2) = P(Y ≤ 10) =
Z 10
0
y −1e−y
(α)
dy =
Z 10
0
y10−1e−y
(10)
dy = 0.542,
with P(X > 2) = 1 − P(X ≤ 2) = 1 − 0.542 = 0.458.
6.57 (a) P(X > 1) = 1 − P(X ≤ 1) = 1 − 10
R 1
0 e−10x dx = e−10 = 0.000045.
(b) μ = β = 1/10 = 0.1.
6.58 Assume that Z(t) = αβt −1, for t > 0. Then we can write f(t) = Z(t)R(t), where
R(t) = ce−R Z(t) dt = ce−R t −1dt = ce− t . From the condition that R(0) = 1, we find
that c = 1. Hence R(t) = e t and f(t) = αβt −1e− t , for t > 0. Since
Z(t) =
f(t)
R(t)
,
where
R(t) = 1 − F(t) = 1 −
Z t
0
αβx −1e− x
dx = 1 +
Z t
0
de− x
= e− t
,
then
Z(t) =
αβt −1e− t
e− t = αβt −1, t > 0.
6.59 μ = np = (1000)(0.49) = 490, σ = √npq =
p
(1000)(0.49)(0.51) = 15.808.
z1 =
481.5 − 490
15.808
= −0.54, z2 =
510.5 − 490
15.808
= 1.3.
P(481.5 < X < 510.5) = P(−0.54 < Z < 1.3) = 0.9032 − 0.2946 = 0.6086.
80 Chapter 6 Some Continuous Probability Distributions
6.60 P(X > 1/4) =
R
∞
1/4 6e−6x dx = −e−6x|∞ 1/4 = e−1.5 = 0.223.
6.61 P(X < 1/2) = 108
R 1/2
0 x2e−6x dx. Letting y = 6x and using Table A.24 we have
P(X < 1/2) = P(Y < 3) =
Z 3
0
y2e−y dy = 0.577.
6.62 Manufacturer A:
P(X ≥ 10000) = P
Z ≥
100000 − 14000
2000
= P(Z ≥ −2) = 0.9772.
Manufacturer B:
P(X ≥ 10000) = P
Z ≥
10000 − 13000
1000
= P(Z ≥ −3) = 0.9987.
Manufacturer B will produce the fewest number of defective rivets.
6.63 Using the normal approximation to the binomial with μ = np = 650 and σ = √npq =
15.0831. So,
P(590 ≤ X ≤ 625) = P(−10.64 < Z < −8.92) ≈ 0.
6.64 (a) μ = β = 100 hours.
(b) P(X ≥ 200) = 0.01
R
∞
200 e−0.01x dx = e−2 = 0.1353.
6.65 (a) μ = 85 and σ = 4. So, P(X < 80) = P(Z < −1.25) = 0.1056.
(b) μ = 79 and σ = 4. So, P(X ≥ 80) = P(Z > 0.25) = 0.4013.
6.66 1/β = 1/5 hours with α = 2 failures and β = 5 hours.
(a) αβ = (2)(5) = 10.
(b) P(X ≥ 12) =
R
∞
12
1
52(2)xe−x/5 dx = 1
25
R
∞
12 xe−x/5 dx =
−x
5 e−x/5 − e−x/5
∞ 1
2
= 0.3084.
6.67 Denote by X the elongation. We have μ = 0.05 and σ = 0.01.
(a) P(X ≥ 0.1) = P
Z ≥ 0.1−0.05
0.01
= P(Z ≥ 5) ≈ 0.
(b) P(X ≤ 0.04) = P
Z ≤ 0.04−0.05
0.01
= P(Z ≤ −1) = 0.1587.
(c) P(0.025 ≤ X ≤ 0.065) = P(−2.5 ≤ Z ≤ 1.5) = 0.9332 − 0.0062 = 0.9270.
6.68 Let X be the error. X ∼ n(x; 0, 4). So,
P(fails) = 1 − P(−10 < X < 10) = 1 − P(−2.25 < Z < 2.25) = 2(0.0122) = 0.0244.
Solutions for Exercises in Chapter 6 81
6.69 Let X be the time to bombing with μ = 3 and σ = 0.5. Then
P(1 ≤ X ≤ 4) = P
1 − 3
0.5 ≤ Z ≤
4 − 3
0.5
= P(−4 ≤ Z ≤ 2) = 0.9772.
P(of an undesirable product) is 1 − 0.9772 = 0.0228. Hence a product is undesirable
is 2.28% of the time.
6.70 α = 2 and β = 100. P(X ≤ 200) = 1
2
R 200
0 xe−x/ dx. Using the incomplete gamma
table and let y = x/β,
R 2
0 ye−y dy = 0.594.
6.71 μ = αβ = 200 hours and σ2 = αβ2 = 20, 000 hours.
6.72 X follows a lognormal distribution.
P(X ≥ 50, 000) = 1 −
ln 50, 000 − 5
2
= 1 − (2.9099) = 1 − 0.9982 = 0.0018.
6.73 The mean of X, which follows a lognormal distribution is μ = E(X) = eμ+ 2/2 = e7.
6.74 μ = 10 and σ = √50.
(a) P(X ≤ 50) = P(Z ≤ 5.66) ≈ 1.
(b) P(X ≤ 10) = 0.5.
(c) The results are very similar.
6.75 (a) Since f(y) ≥ 0 and
R 1
0 10(1 − y)9 dy = − (1 − y)10|1
0 = 1, it is a density function.
(b) P(Y > 0.6) = − (1 − y)10|1
0.6 = (0.4)10 = 0.0001.
(c) α = 1 and β = 10.
(d) μ =
+ = 1
11 = 0.0909.
(e) σ2 =
( + )2( + +1) = (1)(10)
(1+10)2(1+10+1) = 0.006887.
6.76 (a) μ = 1
10
R
∞
0 ze−z/10 dz = − ze−z/10
∞0
+
R
∞
0 e−z/10 dz = 10.
(b) Using integral by parts twice, we get
E(Z2) =
1
10
Z
∞
0
z2e−z/10 dz = 200.
So, σ2 = E(Z2) − μ2 = 200 − (10)2 = 100.
(c) P(Z > 10) = − ez/10
∞ 1
0
=
e−1
=
0.3679.
6.77 This is an exponential distribution with β = 10.
(a) μ = β = 10.
82 Chapter 6 Some Continuous Probability Distributions
(b) σ2 = β2 = 100.
6.78 μ = 0.5 seconds and σ = 0.4 seconds.
(a) P(X > 0.3) = P
Z > 0.3−0.5
0.4
= P(Z > −0.5) = 0.6915.
(b) P(Z > −1.645) = 0.95. So, −1.645 = x−0.5
0.4 yields x = −0.158 seconds. The
negative number in reaction time is not reasonable. So, it means that the normal
model may not be accurate enough.
6.79 (a) For an exponential distribution with parameter β,
P(X > a + b | X > a) =
P(X > a + b)
P(X > a)
=
e−a−b
e−a = e−b = P(X > b).
So, P(it will breakdown in the next 21 days | it just broke down) = P(X > 21) =
e−21/15 = e−1.4 = 0.2466.
(b) P(X > 30) = e−30/15 = e−2 = 0.1353.
6.80 α = 2 and β = 50. So,
P(X ≤ 10) = 100
Z 10
0
x49e−2x50
dx.
Let y = 2x50 with x = (y/2)1/50 and dx = 1
21/50(50)y−49/50 dy.
P(X ≤ 10) =
100
21/50(50)
Z (2)1050
0
y
2
49/50
y−49/50e−y dy =
Z (2)1050
0
e−y dy ≈ 1.
6.81 The density function of a Weibull distribution is
f(y) = αβy −1e− y
, y > 0.
So, for any y ≥ 0,
F(y) =
Z y
0
f(t) dt = αβ
Z y
0
t −1e− t
dt.
Let z = t which yields t = z1/ and dt = 1
z1/ −1 dz. Hence,
F(y) = αβ
Z y
0
z1−1/ 1
β
z1/ −1e− z dz = α
Z y
0
e− z dz = 1 − e− y
.
On the other hand, since de− y = −αβy −1e− y , the above result follows immediately.
Solutions for Exercises in Chapter 6 83
6.82 One of the basic assumptions for the exponential distribution centers around the “lackof-
memory” property for the associated Poisson distribution. Thus the drill bit of
problem 6.80 is assumed to have no punishment through wear if the exponential distribution
applies. A drill bit is a mechanical part that certainly will have significant
wear over time. Hence the exponential distribution would not apply.
6.83 The chi-squared distribution is a special case of the gamma distribution when α = v/2
and β = 2, where v is the degrees of the freedom of the chi-squared distribution.
So, the mean of the chi-squared distribution, using the property from the gamma
distribution, is μ = αβ = (v/2)(2) = v, and the variance of the chi-squared distribution
is σ2 = αβ2 = (v/2)(2)2 = 2v.
6.84 Let X be the length of time in seconds. Then Y = ln(X) follows a normal distribution
with μ = 1.8 and σ = 2.
(a) P(X > 20) = P(Y > ln 20) = P(Z > (ln 20 − 1.8)/2) = P(Z > 0.60) = 0.2743.
P(X > 60) = P(Y > ln 60) = P(Z > (ln 60 − 1.8)/2) = P(Z > 1.15) = 0.1251.
(b) The mean of the underlying normal distribution is e1.8+4/2 = 44.70 seconds. So,
P(X < 44.70) = P(Z < (ln 44.70 − 1.8)/2) = P(Z < 1) = 0.8413.
Chapter 7
Functions of Random Variables
7.1 From y = 2x − 1 we obtain x = (y + 1)/2, and given x = 1, 2, and 3, then
g(y) = f[(y + 1)/2] = 1/3, for y = 1, 3, 5.
7.2 From y = x2, x = 0, 1, 2, 3, we obtain x = √y,
g(y) = f(√y) =
3
√y
2
5
√y
3
5
3−√y
, fory = 0, 1, 4, 9.
7.3 The inverse functions of y1 = x1 + x2 and y2 = x1 − x2 are x1 = (y1 + y2)/2 and
x2 = (y1 − y2)/2. Therefore,
g(y1, y2) =
2
y1+y2
2 , y1−y2
2 , 2 − y1
1
4
(y1+y2)/2
1
3
(y1−y2)/2
5
12
2−y1
,
where y1 = 0, 1, 2, y2 = −2,−1, 0, 1, 2, y2 ≤ y1 and y1 + y2 = 0, 2, 4.
7.4 Let W = X2. The inverse functions of y = x1x2 and w = x2 are x1 = y/w and x2 = w,
where y/w = 1, 2. Then
g(y,w) = (y/w)(w/18) = y/18, y = 1, 2, 3, 4, 6; w = 1, 2, 3, y/w = 1, 2.
In tabular form the joint distribution g(y,w) and marginal h(y) are given by
y
g(y,w) 1 2 3 4 6
1 1/18 2/18
w 2 2/18 4/18
3 3/18 6/18
h(y) 1/18 2/9 1/6 2/9 1/3
85
86 Chapter 7 Functions of Random Variables
The alternate solutions are:
P(Y = 1) = f(1, 1) = 1/18,
P(Y = 2) = f(1, 2) + f(2, 1) = 2/18 + 2/18 = 2/9,
P(Y = 3) = f(1, 3) = 3/18 = 1/6,
P(Y = 4) = f(2, 2) = 4/18 = 2/9,
P(Y = 6) = f(2, 3) = 6/18 = 1/3.
7.5 The inverse function of y = −2 ln x is given by x = e−y/2 from which we obtain
|J| = | − e−y/2/2| = e−y/2/2. Now,
g(y) = f(ey/2)|J| = e−y/2/2, y > 0,
which is a chi-squared distribution with 2 degrees of freedom.
7.6 The inverse function of y = 8x3 is x = y1/3/2, for 0 < y < 8 from which we obtain
|J| = y−2/3/6. Therefore,
g(y) = f(y1/3/2)|J| = 2(y1/3/2)(y−2/3/6) =
1
6
y−1/3, 0 < y < 8.
7.7 To find k we solve the equation k
R
∞
0 v2e−bv2 dv = 1. Let x = bv2, then dx = 2bv dv
and dv = x−1/2
2√b
dx. Then the equation becomes
k
2b3/2
Z
∞
0
x3/2−1e−x dx = 1, or
k(3/2)
2b3/2 = 1.
Hence k = 4b3/2
(1/2) .
Now the inverse function of w = mv2/2 is v =
p
2w/m, for w > 0, from which we
obtain |J| = 1/√2mw. It follows that
g(w) = f(
p
2w/m)|J| =
4b3/2
(1/2)
(2w/m)e−2bw/m =
1
(m/2b)3/2(3/2)
w3/2−1e−(2b/m)w,
for w > 0, which is a gamma distribution with α = 3/2 and β = m/2b.
7.8 (a) The inverse of y = x2 is x = √y, for 0 < y < 1, from which we obtain |J| = 1/2√y.
Therefore,
g(y) = f(√y)|J| = 2(1 − √y)/2√y = y−1/2 − 1, 0 < y < 1.
(b) P(Y < 1) =
R 1
0 (y−1/2 − 1) dy = (2y1/2 − y)
1
0 = 0.5324.
7.9 (a) The inverse of y = x + 4 is x = y − 4, for y > 4, from which we obtain |J| = 1.
Therefore,
g(y) = f(y − 4)|J| = 32/y3, y > 4.
Solutions for Exercises in Chapter 7 87
(b) P(Y > 8) = 32
R
∞
8 y−3 dy = − 16y−2|∞8 = 1
4 .
7.10 (a) Let W = X. The inverse functions of z = x + y and w = x are x = w and
y = z − w, 0 < w < z, 0 < z < 1, from which we obtain
J =
@x
@w
@x
@z
@y
@w
@y
@z
=
1 0
−1 1
= 1.
Then g(w, z) = f(w, z − w)|J| = 24w(z − w), for 0 < w < z and 0 < z < 1. The
marginal distribution of Z is
f1(z) =
Z 1
0
24(z − w)w dw = 4z3, 0 < z < 1.
(b) P(1/2 < Z < 3/4) = 4
R 3/4
1/2 z3 dz = 65/256.
7.11 The amount of kerosene left at the end of the day is Z = Y − X. Let W = Y . The
inverse functions of z = y − x and w = y are x = w − z and y = w, for 0 < z < w and
0 < w < 1, from which we obtain
J =
@x
@w
@x
@z
@y
@w
@y
@z
=
1 −1
1 0
= 1.
Now,
g(w, z) = g(w − z,w) = 2, 0 < z < w, 0 < w < 1,
and the marginal distribution of Z is
h(z) = 2
Z 1
z
dw = 2(1 − z), 0 < z < 1.
7.12 Since X1 and X2 are independent, the joint probability distribution is
f(x1, x2) = f(x1)f(x2) = e−(x1+x2), x1 > 0, x2 > 0.
The inverse functions of y1 = x1 + x2 and y2 = x1/(x1 + x2) are x1 = y1y2 and
x2 = y1(1 − y2), for y1 > 0 and 0 < y2 < 1, so that
J =
∂x1/∂y1 ∂x1/∂y2
∂x2/∂y1 ∂x2/∂y2
=
y2 y1
1 − y2 −y1
= −y1.
Then, g(y1, y2) = f(y1y2, y1(1−y2))|J| = y1e−y1 , for y1 > 0 and 0 < y2 < 1. Therefore,
g(y1) =
Z 1
0
y1e−y1 dy2 = y1e−y1 , y1 > 0,
and
g(y2) =
Z
∞
0
y1e−y1 dy1 = (2) = 1, 0 < y2 < 1.
Since g(y1, y2) = g(y1)g(y2), the random variables Y1 and Y2 are independent.
88 Chapter 7 Functions of Random Variables
7.13 Since I and R are independent, the joint probability distribution is
f(i, r) = 12ri(1 − i), 0 < i < 1, 0 < r < 1.
Let V = R. The inverse functions of w = i2r and v = r are i =
p
w/v and r = v, for
w < v < 1 and 0 < w < 1, from which we obtain
J =
∂i/∂w ∂i/∂v
∂r/∂w ∂r/∂v
=
1
2√vw
.
Then,
g(w, v) = f(
p
w/v, v)|J| = 12v
p
w/v(1 −
p
w/v)
1
2√vw
= 6(1 −
p
w/v),
for w < v < 1 and 0 < w < 1, and the marginal distribution of W is
h(w) = 6
Z 1
w
(1 −
p
w/v) dv = 6 (v − 2√wv)
v
=
1
v=w = 6 + 6w − 12√w, 0 < w < 1.
7.14 The inverse functions of y = x2 are given by x1 = √y and x2 = −√y from which we
obtain J1 = 1/2√y and J2 = 1/2√y. Therefore,
g(y) = f(√y)|J1| + f(−√y)|J2| =
1 + √y
2 ·
1
2√y
+
1 − √y
2 ·
1
2√y
= 1/2√y,
for 0 < y < 1.
7.15 The inverse functions of y = x2 are x1 = √y, x2 = −√y for 0 < y < 1 and x1 = √y
for 0 < y < 4. Now |J1| = |J2| = |J3| = 1/2√y, from which we get
g(y) = f(√y)|J1| + f(−√y)|J2| =
2(√y + 1)
9 ·
1
2√y
+
2(−√y + 1)
9 ·
1
2√y
=
2
9√y
,
for 0 < y < 1 and
g(y) = f(√y)|J3| =
2(√y + 1)
9 ·
1
2√y
=
√y + 1
9√y
, for 1 < y < 4.
7.16 Using the formula we obtain
μ
′
r = E(Xr) =
Z
∞
0
xr ·
x −1e−x/
β (α)
dx =
β +r(α + r)
β (α)
Z
∞
0
x +r−1e−x/
β +r(α + r)
dx
=
βr(α + r)
(α)
,
since the second integrand is a gamma density with parametersα + r and β.
Solutions for Exercises in Chapter 7 89
7.17 The moment-generating function of X is
MX(t) = E(etX) =
1
k
Xk
x=1
etx =
et(1 − ekt)
k(1 − et)
,
by summing the geometric series of k terms.
7.18 The moment-generating function of X is
MX(t) = E(etX) = p
∞X
x=1
etxqx−1 =
p
q
∞X
x=1
(etq)x =
pet
1 − qet ,
by summing an infinite geometric series. To find out the moments, we use
μ = M
′
X(0) =
(1 − qet)pet + pqe2t
(1 − qet)2
t=0
=
(1 − q)p + pq
(1 − q)2 =
1
p
,
and
μ
′
2 = M
′′
X(0) =
(1 − qet)2pet + 2pqe2t(1 − qet)
(1 − qet)4
t=0
=
2 − p
p2 .
So, σ2 = μ′
2 − μ2 = q
p2 .
7.19 The moment-generating function of a Poisson random variable is
MX(t) = E(etX) =
∞X
x=0
etxe−μμx
x!
= e−μ
∞X
x=0
(μet)x
x!
= e−μeμet
= eμ(et−1).
So,
μ = M
′
X(0) = μ eμ(et−1)+t
t=0
= μ,
μ
′
2 = M
′′
X(0) = μeμ(et−1)+t(μet + 1)
t=0
= μ(μ + 1),
and
σ2 = μ
′
2 − μ2 = μ(μ + 1) − μ2 = μ.
7.20 From MX(t) = e4(et−1) we obtain μ = 6, σ2 = 4, and σ = 2. Therefore,
P(μ − 2σ < X < μ + 2σ) = P(0 < X < 8) =
X7
x=1
p(x; 4) = 0.9489 − 0.0183 = 0.9306.
90 Chapter 7 Functions of Random Variables
7.21 Using the moment-generating function of the chi-squared distribution, we obtain
μ = M
′
X(0) = v(1 − 2t)−v/2−1
t=0 = v,
μ
′
2 = M
′′
X(0) = v(v + 2) (1 − 2t)−v/2−2
t=0 = v(v + 2).
So, σ2 = μ′
2 − μ2 = v(v + 2) − v2 = 2v.
7.22
MX(t) =
Z
∞
−∞
etxf(x) dx =
Z
∞
−∞
1 + tx +
t2x2
2!
+ · · · +
trxr
r!
+ · · ·
f(x) dx
=
Z
∞
−∞
f(x) dx + t
Z
∞
−∞
xf(x) dx +
t2
2
Z
∞
−∞
x2f(x) dx
+ · · · +
tr
r!
Z
∞
−∞
xrf(x) dx + · · · = 1 + μt + μ
′
1
t2
2!
+ · · · + μ
′
r
tr
r!
+ · · · .
7.23 The joint distribution of X and Y is fX,Y (x, y) = e−x−y for x > 0 and y > 0. The
inverse functions of u = x + y and v = x/(x + y) are x = uv and y = u(1 − v) with
J =
v u
1 − v −u
= u for u > 0 and 0 < v < 1. So, the joint distribution of U and V is
fU,V (u, v) = ue−uv · e−u(1−v) = ue−u,
for u > 0 and 0 < v < 1.
(a) fU (u) =
R 1
0 ue−u dv = ue−u for u > 0, which is a gamma distribution with
parameters 2 and 1.
(b) fV (v) =
R
∞
0 ue−u du = 1 for 0 < v < 1. This is a uniform (0,1) distribution.
Chapter 8
Fundamental Sampling Distributions
and Data Descriptions
8.1 (a) Responses of all people in Richmond who have telephones.
(b) Outcomes for a large or infinite number of tosses of a coin.
(c) Length of life of such tennis shoes when worn on the professional tour.
(d) All possible time intervals for this lawyer to drive from her home to her office.
8.2 (a) Number of tickets issued by all state troopers in Montgomery County during the
Memorial holiday weekend.
(b) Number of tickets issued by all state troopers in South Carolina during the Memorial
holiday weekend.
8.3 (a) ¯x = 2.4.
(b) ¯x = 2.
(c) m = 3.
8.4 (a) ¯x = 8.6 minutes.
(b) ¯x = 9.5 minutes.
(c) Mode are 5 and 10 minutes.
8.5 (a) ¯x = 3.2 seconds.
(b) ¯x = 3.1 seconds.
8.6 (a) ¯x = 35.7 grams.
(b) ¯x = 32.5 grams.
(c) Mode=29 grams.
8.7 (a) ¯x = 53.75.
91
92 Chapter 8 Fundamental Sampling Distributions and Data Descriptions
(b) Modes are 75 and 100.
8.8 ¯x = 22.2 days, ˜x = 14 days and m = 8 days. ˜x is the best measure of the center of the
data. The mean should not be used on account of the extreme value 95, and the mode
is not desirable because the sample size is too small.
8.9 (a) Range = 15 − 5 = 10.
(b) s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (10)(838)−862
(10)(9) = 10.933. Taking the square root, we have
s = 3.307.
8.10 (a) Range = 4.3 − 2.3 = 2.0.
(b) s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (9)(96.14)−28.82
(9)(8) = 0.498.
8.11 (a) s2 = 1
n−1
Pn
x=1
(xi − ¯x)2 = 1
14 [(2 − 2.4)2 + (1 − 2.4)2 + · · · + (2 − 2.4)2] = 2.971.
(b) s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (15)(128)−362
(15)(14) = 2.971.
8.12 (a) ¯x = 11.69 milligrams.
(b) s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (8)(1168.21)−93.52
(8)(7) = 10.776.
8.13 s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (2)(148.55)−53.32
(20)(19) = 0.342 and hence s = 0.585.
8.14 (a) Replace Xi in S2 by Xi + c for i = 1, 2, . . . , n. Then ¯X becomes ¯X + c and
S2 =
1
n − 1
Xn
i=1
[(Xi + c) − ( ¯X + c)]2 =
1
n − 1
Xn
i=1
(Xi − ¯X )2.
(b) Replace Xi by cXi in S2 for i = 1, 2, . . . , n. Then ¯X becomes c ¯X and
S2 =
1
n − 1
Xn
i=1
(cXi − c ¯X )2 =
c2
n − 1
Xn
i=1
(Xi − ¯X )2.
8.15 s2 =
n
n P
i
=
1
x2
i −(
n P
i
=
1
xi)2
n(n−1) = (6)(207)−332
(6)(5) = 5.1.
(a) Multiplying each observation by 3 gives s2 = (9)(5.1) = 45.9.
(b) Adding 5 to each observation does not change the variance. Hence s2 = 5.1.
8.16 Denote by D the difference in scores.
Solutions for Exercises in Chapter 8 93
(a) ¯D = 25.15.
(b) ˜D = 31.00.
8.17 z1 = −1.9, z2 = −0.4. Hence,
P(μ ¯X − 1.9σ ¯X < ¯X < μ ¯X − 0.4σ ¯X ) = P(−1.9 < Z < −0.4) = 0.3446 − 0.0287 = 0.3159.
8.18 n = 54, μ ¯X = 4, σ2
¯X
= σ2/n = (8/3)/54 = 4/81 with σ ¯X = 2/9. So,
z1 = (4.15 − 4)/(2/9) = 0.68, and z2 = (4.35 − 4)/(2/9) = 1.58,
and
P(4.15 < ¯X < 4.35) = P(0.68 < Z < 1.58) = 0.9429 − 0.7517 = 0.1912.
8.19 (a) For n = 64, σ ¯X = 5.6/8 = 0.7, whereas for n = 196, σ ¯X = 5.6/14 = 0.4.
Therefore, the variance of the sample mean is reduced from 0.49 to 0.16 when the
sample size is increased from 64 to 196.
(b) For n = 784, σ ¯X = 5.6/28 = 0.2, whereas for n = 49, σ ¯X = 5.6/7 = 0.8.
Therefore, the variance of the sample mean is increased from 0.04 to 0.64 when
the sample size is decreased from 784 to 49.
8.20 n = 36, σ ¯X = 2. Hence σ = √nσ ¯X = (6)(2) = 12. If σ ¯X = 1.2, then 1.2 = 12/√n and
n = 100.
8.21 μ ¯X = μ = 240, σ ¯X = 15/√40 = 2.372. Therefore, μ ¯X ±2σ ¯X = 240±(2)(2.372) or from
235.257 to 244.743, which indicates that a value of x = 236 milliliters is reasonable
and hence the machine needs not be adjusted.
8.22 (a) μ ¯X = μ = 174.5, σ ¯X = σ/√n = 6.9/5 = 1.38.
(b) z1 = (172.45 − 174.5)/1.38 = −1.49, z2 = (175.85 − 174.5)/1.38 = 0.98. So,
P(172.45 < ¯X < 175.85) = P(−1.49 < Z < 0.98) = 0.8365 − 0.0681 = 0.7684.
Therefore, the number of sample means between 172.5 and 175.8 inclusive is
(200)(0.7684) = 154.
(c) z = (171.95 − 174.5)/1.38 = −1.85. So,
P( ¯X < 171.95) = P(Z < −1.85) = 0.0322.
Therefore, about (200)(0.0322) = 6 sample means fall below 172.0 centimeters.
8.23 (a) μ =
P
xf(x) = (4)(0.2) + (5)(0.4) + (6)(0.3) + (7)(0.1) = 5.3, and
σ2 =
P
(x − μ)2f(x) = (4 − 5.3)2(0.2) + (5 − 5.3)2(0.4) + (6 − 5.3)2(0.3) + (7 −
5.3)2(0.1) = 0.81.
94 Chapter 8 Fundamental Sampling Distributions and Data Descriptions
(b) With n = 36, μ ¯X = μ = 5.3 and σ ¯X = σ2/n = 0.81/36 = 0.0225.
(c) n = 36, μ ¯X = 5.3, σ ¯X = 0.9/6 = 0.15, and z = (5.5 − 5.3)/0.15 = 1.33. So,
P( ¯X < 5.5) = P(Z < 1.33) = 0.9082.
8.24 n = 36, μ ¯X = 40, σ ¯X = 2/6 = 1/3 and z = (40.5 − 40)/(1/3) = 1.5. So,
P
X36
i=1
Xi > 1458
!
= P( ¯X > 40.5) = P(Z > 1.5) = 1 − 0.9332 = 0.0668.
8.25 (a) P(6.4 < ¯X < 7.2) = P(−1.8 < Z < 0.6) = 0.6898.
(b) z = 1.04, ¯x = z(σ/√n) + μ = (1.04)(1/3) + 7 = 7.35.
8.26 n = 64, μ ¯X = 3.2, σ ¯X = σ/√n = 1.6/8 = 0.2.
(a) z = (2.7 − 3.2)/0.2 = −2.5, P( ¯X < 2.7) = P(Z < −2.5) = 0.0062.
(b) z = (3.5 − 3.2)/0.2 = 1.5, P( ¯X > 3.5) = P(Z > 1.5) = 1 − 0.9332 = 0.0668.
(c) z1 = (3.2 − 3.2)/0.2 = 0, z2 = (3.4 − 3.2)/0.2 = 1.0,
P(3.2 < ¯X < 3.4) = P(0 < Z < 1.0) = 0.9413 − 0.5000 = 0.3413.
8.27 n = 50, ¯x = 0.23 and σ = 0.1. Now, z = (0.23 − 0.2)/(0.1/√50) = 2.12; so
P( ¯X ≥ 0.23) = P(Z ≥ 2.12) = 0.0170.
Hence the probability of having such observations, given the mean μ = 0.20, is small.
Therefore, the mean amount to be 0.20 is not likely to be true.
8.28 μ1−μ2 = 80−75 = 5, σ ¯X1− ¯X2 =
p
25/25 + 9/36 = 1.118, z1 = (3.35−5)/1.118 = −1.48
and z2 = (5.85 − 5)/1.118 = 0.76. So,
P(3.35 < ¯X1 − ¯X2 < 5.85) = P(−1.48 < Z < 0.76) = 0.7764 − 0.0694 = 0.7070.
8.29 μ ¯X1− ¯X2 = 72 − 28 = 44, σ ¯X1− ¯X2 =
p
100/64 + 25/100 = 1.346 and z = (44.2 − 44)/1.346 = 0.15. So, P( ¯X1 − ¯X2 < 44.2) = P(Z < 0.15) = 0.5596.
8.30 μ1 − μ2 = 0, σ ¯X1− ¯X2 = 50
p
1/32 + 1/50 = 11.319.
(a) z1 = −20/11.319 = −1.77, z2 = 20/11.319 = 1.77, so
P(| ¯X1 − ¯X2| > 20) = 2P(Z < −1.77) = (2)(0.0384) = 0.0768.
(b) z1 = 5/11.319 = 0.44 and z2 = 10/11.319 = 0.88. So,
P(−10 < ¯X1 − ¯X2 < −5) + P(5 < ¯X1 − ¯X2 < 10) = 2P(5 < ¯X1 − ¯X2 < 10) =
2P(0.44 < Z < 0.88) = 2(0.8106 − 0.6700) = 0.2812.
8.31 The normal quantile-quantile plot is shown as
Solutions for Exercises in Chapter 8 95
−2 −1 0 1 2
700 800 900 1000 1100 1200 1300
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
8.32 (a) If the two population mean drying times are truly equal, the probability that the
difference of the two sample means is 1.0 is 0.0013, which is very small. This means
that the assumption of the equality of the population means are not reasonable.
(b) If the experiment was run 10,000 times, there would be (10000)(0.0013) = 13
experiments where ¯XA − ¯XB would be at least 1.0.
8.33 (a) n1 = n2 = 36 and z = 0.2/
p
1/36 + 1/36 = 0.85. So,
P( ¯XB − ¯XA ≥ 0.2) = P(Z ≥ 0.85) = 0.1977.
(b) Since the probability in (a) is not negligible, the conjecture is not true.
8.34 The normal quantile-quantile plot is shown as
−2 −1 0 1 2
6.65 6.70 6.75 6.80
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
96 Chapter 8 Fundamental Sampling Distributions and Data Descriptions
8.35 (a) When the population equals the limit, the probability of a sample mean exceeding
the limit would be 1/2 due the symmetry of the approximated normal distribution.
(b) P( ¯X ≥ 7960 | μ = 7950) = P(Z ≥ (7960 − 7950)/(100/√25)) = P(Z ≥ 0.5) =
0.3085. No, this is not very strong evidence that the population mean of the
process exceeds the government limit.
8.36 (a) σ ¯XA− ¯XB =
q
52
30 + 52
30 = 1.29 and z = 4−0
1.29 = 3.10. So,
P( ¯XA − ¯XB > 4 | μA = μB) = P(Z > 3.10) = 0.0010.
Such a small probability means that the difference of 4 is not likely if the two
population means are equal.
(b) Yes, the data strongly support alloy A.
8.37 Since the probability that ¯X ≤ 775 is 0.0062, given that μ = 800 is true, it suggests
that this event is very rare and it is very likely that the claim of μ = 800 is not true.
On the other hand, if μ is truly, say, 760, the probability
P( ¯X ≤ 775 | μ = 760) = P(Z ≤ (775 − 760)/(40/√16)) = P(Z ≤ 1.5) = 0.9332,
which is very high.
8.38 Define Wi = lnXi for i = 1, 2, . . . . Using the central limit theorem, Z = ( ¯W − μW1)/(σW1/√n) ∼ n(z; 0, 1). Hence ¯W follows, approximately, a normal distribution
when n is large. Since
¯W
=
1
n
Xn
i=1
ln(Xi) =
1
n
ln
Yn
i=1
Xi
!
=
1
n
ln(Y ),
then it is easily seen that Y follows, approximately, a lognormal distribution.
8.39 (a) 27.488.
(b) 18.475.
(c) 36.415.
8.40 (a) 16.750.
(b) 30.144.
(c) 26.217.
8.41 (a) χ2
= χ2
0.99 = 0.297.
(b) χ2
= χ2
0.025 = 32.852.
(c) χ2
0.05 = 37.652. Therefore, α = 0.05−0.045 = 0.005. Hence, χ2
= χ2
0.005 = 46.928.
8.42 (a) χ2
= χ2
0.01 = 38.932.
Solutions for Exercises in Chapter 8 97
(b) χ2
= χ2
0.05 = 12.592.
(c) χ2
0.01 = 23.209 and χ2
0.025 = 20.483 with α = 0.01 + 0.015 = 0.025.
8.43 (a) P(S2 > 9.1) = P
(n−1)S2
2 > (24)(9.1)
6
= P(χ2 > 36.4) = 0.05.
(b) P(3.462 < S2 < 10.745) = P
(24)(3.462)
6 < (n−1)S2
2 < (24)(10.745)
6
= P(13.848 < χ2 < 42.980) = 0.95 − 0.01 = 0.94.
8.44 χ2 = (19)(20)
8 = 47.5 while χ2
0.01 = 36.191. Conclusion values are not valid.
8.45 Since (n−1)S2
2 is a chi-square statistic, it follows that
σ2
(n−1)S2/ 2 =
(n − 1)2
σ4 σ2
S2 = 2(n − 1).
Hence, σ2
S2 = 2 4
n−1 , which decreases as n increases.
8.46 (a) 2.145.
(b) −1.372.
(c) −3.499.
8.47 (a) P(T < 2.365) = 1 − 0.025 = 0.975.
(b) P(T > 1.318) = 0.10.
(c) P(T < 2.179) = 1 − 0.025 = 0.975, P(T < −1.356) = P(T > 1.356) = 0.10.
Therefore, P(−1.356 < T < 2.179) = 0.975 − 0.010 = 0.875.
(d) P(T > −2.567) = 1 − P(T > 2.567) = 1 − 0.01 = 0.99.
8.48 (a) Since t0.01 leaves an area of 0.01 to the right, and −t0.005 an area of 0.005 to the
left, we find the total area to be 1 − 0.01 − 0.005 = 0.985 between −t0.005 and
t0.01. Hence, P(−t0.005 < T < t0.01) = 0.985.
(b) Since −t0.025 leaves an area of 0.025 to the left, the desired area is 1−0.025 = 0.975.
That is, P(T > −t0.025) = 0.975.
8.49 (a) From Table A.4 we note that 2.069 corresponds to t0.025 when v = 23. Therefore,
−t0.025 = −2.069 which means that the total area under the curve to the left of
t = k is 0.025 + 0.965 = 0.990. Hence, k = t0.01 = 2.500.
(b) From Table A.4 we note that 2.807 corresponds to t0.005 when v = 23. Therefore
the total area under the curve to the right of t = k is 0.095+0.005 = 0.10. Hence,
k = t0.10 = 1.319.
(c) t0.05 = 1.714 for 23 degrees of freedom.
98 Chapter 8 Fundamental Sampling Distributions and Data Descriptions
8.50 From Table A.4 we find t0.025 = 2.131 for v = 15 degrees of freedom. Since the value
t =
27.5 − 30
5/4
= −2.00
falls between −2.131 and 2.131, the claim is valid.
8.51 t = (24 − 20)/(4.1/3) = 2.927, t0.01 = 2.896 with 8 degrees of freedom. Conclusion:
no, μ > 20.
8.52 ¯x = 0.475, s2 = 0.0336 and t = (0.475 − 0.5)/0.0648 = −0.39. Hence
P( ¯X < 0.475) = P(T < −0.39) ≈ 0.35.
So, the result is inconclusive.
8.53 (a) 2.71.
(b) 3.51.
(c) 2.92.
(d) 1/2.11 = 0.47.
(e) 1/2.90 = 0.34.
8.54 s2
1 = 10.441 and s2
2 = 1.846 which gives f = 5.66. Since, from Table A.6, f0.05(9, 7) =
3.68 and f0.01(9, 7) = 6.72, the probability of P(F > 5.66) should be between 0.01 and
0.05, which is quite small. Hence the variances may not be equal. Furthermore, if a
computer software can be used, the exact probability of F > 5.66 can be found 0.0162,
or if two sides are considered, P(F < 1/5.66) + P(F > 5.66) = 0.026.
8.55 s2
1 = 15750 and s2
2 = 10920 which gives f = 1.44. Since, from Table A.6, f0.05(4, 5) =
5.19, the probability of F > 1.44 is much bigger than 0.05, which means that the two
variances may be considered equal. The actual probability of F > 1.44 is 0.3436 and
P(F < 1/1.44) + P(F > 1.44) = 0.7158.
8.56 The box-and-whisker plot is shown below.
5 10 15 20
Box−and−Whisker Plot
Solutions for Exercises in Chapter 8 99
The sample mean = 12.32 and the sample standard deviation = 6.08.
8.57 The moment-generating function for the gamma distribution is given by
MX(t) = E(etX) =
1
β (α)
Z
∞
0
etxx −1e−x/ dx
=
1
β (1/β − t)
1
(1/β − t)− (α)
Z
∞
0
x −1e−x( 1
−t) dx
=
1
(1 − βt)
Z
∞
0
x −1e−x/(1/ −t)−1
[(1/β − t)−1] (α)
dx =
1
(1 − βt) ,
for t < 1/β, since the last integral is one due to the integrand being a gamma density
function. Therefore, the moment-generating function of an exponential distribution,
by substituting α to 1, is given by MX(t) = (1−θt)−1. Hence, the moment-generating
function of Y can be expressed as
MY (t) = MX1(t)MX2 (t) · · ·MXn(t) =
Yn
i=1
(1 − θt)−1 = (1 − θt)−n,
which is seen to be the moment-generating function of a gamma distribution with
α = n and β = θ.
8.58 The variance of the carbon monoxide contents is the same as the variance of the coded
measurements. That is, s2 = (15)(199.94)−392
(15)(14) = 7.039, which results in s = 2.653.
8.59 P
S2
1
S2
2
< 4.89
= P
S2
1/ 2
S2
2/ 2 < 4.89
= P(F < 4.89) = 0.99, where F has 7 and 11
degrees of freedom.
8.60 s2 = 114, 700, 000.
8.61 Let X1 and X2 be Poisson variables with parameters λ1 = 6 and λ2 = 6 representing the
number of hurricanes during the first and second years, respectively. Then Y = X1+X2
has a Poisson distribution with parameter λ = λ1 + λ2 = 12.
(a) P(Y = 15) = e−121215
15! = 0.0724.
(b) P(Y ≤ 9) =
P9
y=0
e−1212y
y! = 0.2424.
8.62 Dividing each observation by 1000 and then subtracting 55 yields the following data:
−7, −2, −10, 6, 4, 1, 8, −6, −2, and −1. The variance of this coded data is
(10)(311) − (−9)2
(10)(9)
= 33.656.
Hence, with c = 1000, we have
s2 = (1000)2(33.656) = 33.656 × 106,
and then s = 5801 kilometers.
100 Chapter 8 Fundamental Sampling Distributions and Data Descriptions
8.63 The box-and-whisker plot is shown below.
5 10 15 20 25
Box−and_Whisker Plot
The sample mean is 2.7967 and the sample standard deviation is 2.2273.
8.64 P
S2
1
S2
2
> 1.26
= P
S2
1/ 2
1
S2
2/ 2
2
> (15)(1.26)
10
= P(F > 1.89) ≈ 0.05, where F has 24 and 30
degrees of freedom.
8.65 No outliers.
8.66 The value 32 is a possible outlier.
8.67 μ = 5,000 psi, σ = 400 psi, and n = 36.
(a) Using approximate normal distribution (by CLT),
P(4800 < ¯X < 5200) = P
4800 − 5000
400/√36
< Z <
5200 − 5000
400/√36
= P(−3 < Z < 3) = 0.9974.
(b) To find a z such that P(−z < Z < z) = 0.99, we have P(Z < z) = 0.995, which
results in z = 2.575. Hence, by solving 2.575 = 5100−5000
400/√n we have n ≥ 107. Note
that the value n can be affected by the z values picked (2.57 or 2.58).
8.68 ¯x = 54,100 and s = 5801.34. Hence
t =
54100 − 53000
5801.34/√10
= 0.60.
So, P( ¯X ≥ 54, 100) = P(T ≥ 0.60) is a value between 0.20 and 0.30, which is not a
rare event.
8.69 nA = nB = 20, ¯xA = 20.50, ¯xB = 24.50, and σA = σB = 5.
(a) P( ¯XA − ¯XB ≥ 4.0 | μA = μB) = P(Z > (24.5 − 20.5)/
p
52/20 + 52/20)
= P(Z > 4.5/(5/√10)) = P(Z > 2.85) = 0.0022.
Solutions for Exercises in Chapter 8 101
(b) It is extremely unlikely that μA = μB.
8.70 (a) nA = 30, ¯xA = 64.5% and σA = 5%. Hence,
P( ¯XA ≤ 64.5 | μA = 65) = P(Z < (64.5 − 65)/(5/√30)) = P(Z < −0.55)
= 0.2912.
There is no evidence that the μA is less than 65%.
(b) nB = 30, ¯xB = 70% and σB = 5%. It turns out σ ¯XB− ¯XA =
q
52
30 + 52
30 = 1.29%.
Hence,
P( ¯XB − ¯XA ≥ 5.5 | μA = μB) = P
Z ≥
5.5
1.29
= P(Z ≥ 4.26) ≈ 0.
It does strongly support that μB is greater than μA.
(c) i) Since σ ¯XB = 5 √30
= 0.9129, ¯XB ∼ n(x; 65%, 0.9129%).
ii) ¯XA − ¯XB ∼ n(x; 0, 1.29%).
iii)
¯X
A− ¯XB
√2/30 ∼ n(z; 0, 1).
8.71 P( ¯XB ≥ 70) = P
Z ≥ 70−65
0.9129
= P(Z ≥ 5.48) ≈ 0.
8.72 It is known, from Table A.3, that P(−1.96 < Z < 1.96) = 0.95. Given μ = 20 and
σ = √9 = 3, we equate 1.96 = 20.1−20
3/√n to obtain n =
(3)(1.96)
0.1
2
= 3457.44 ≈ 3458.
8.73 It is known that P(−2.575 < Z < 2.575) = 0.99. Hence, by equating 2.575 = 0.05
1/√n, we
obtain n =
2.575
0.05
2 = 2652.25 ≈ 2653.
8.74 μ = 9 and σ = 1. Then
P(9 − 1.5 < X < 9 + 1.5) = P(7.5 < X < 10.5) = P(−1.5 < Z < 1.5)
= 0.9322 − 0.0668 = 0.8654.
Thus the proportion of defective is 1 − 0.8654 = 0.1346. To meet the specifications
99% of the time, we need to equate 2.575 = 1.5
, since P(−2.575 < Z < 2.575) = 0.99.
Therefore, σ = 1.5
2.575 = 0.5825.
8.75 With the 39 degrees of freedom,
P(S2 ≤ 0.188 | σ2 = 1.0) = P(χ2 ≤ (39)(0.188)) = P(χ2 ≤ 7.332) ≈ 0,
which means that it is impossible to observe s2 = 0.188 with n = 40 for σ2 = 1.
Note that Table A.5 does not provide any values for the degrees of freedom to be larger
than 30. However, one can deduce the conclusion based on the values in the last line
of the table. Also, computer software gives the value of 0.
Chapter 9
One- and Two-Sample Estimation
Problems
9.1 From Example 9.1 on page 271, we know that E(S2) = σ2. Therefore,
E(S′2) = E
n − 1
n
S2
=
n − 1
n
E(S2) =
n − 1
n
σ2.
9.2 (a) E(X) = np; E( ˆ P) = E(X/n) = E(X)/n = np/n = p.
(b) E(P′) = E(X)+√n/2
n+√n = np+√n/2
n+√n 6= p.
9.3 lim
n→∞
np+√n/2
n+√n = lim
n→∞
p+1/2√n
1+1/√n = p.
9.4 n = 30, ¯x = 780, and σ = 40. Also, z0.02 = 2.054. So, a 96% confidence interval for the
population mean can be calculated as
780 − (2.054)(40/√30) < μ < 780 + (2.054)(40/√30),
or 765 < μ < 795.
9.5 n = 75, ¯x = 0.310, σ = 0.0015, and z0.025 = 1.96. A 95% confidence interval for the
population mean is
0.310 − (1.96)(0.0015/√75) < μ < 0.310 + (1.96)(0.0015/√75),
or 0.3097 < μ < 0.3103.
9.6 n = 50, ¯x = 174.5, σ = 6.9, and z0.01 = 2.33.
(a) A 98% confidence interval for the population mean is
174.5 − (2.33)(6.9/√50) < μ < 174.5 + (2.33)(6.9/√50), or 172.23 < μ < 176.77.
(b) e < (2.33)(6.9)/√50 = 2.27.
103
104 Chapter 9 One- and Two-Sample Estimation Problems
9.7 n = 100, ¯x = 23, 500, σ = 3900, and z0.005 = 2.575.
(a) A 99% confidence interval for the population mean is
23, 500 − (2.575)(3900/10) < μ < 23, 500 + (2.575)(3900/10), or
22, 496 < μ < 24, 504.
(b) e < (2.575)(3900/10) = 1004.
9.8 n = [(2.05)(40)/10]2 = 68 when rounded up.
9.9 n = [(1.96)(0.0015)/0.0005]2 = 35 when rounded up.
9.10 n = [(1.96)(40)/15]2 = 28 when rounded up.
9.11 n = [(2.575)(5.8)/2]2 = 56 when rounded up.
9.12 n = 20, ¯x = 11.3, s = 2.45, and t0.025 = 2.093 with 19 degrees of freedom. A 95%
confidence interval for the population mean is
11.3 − (2.093)(2.45/√20) < μ < 11.3 + (2.093)(2.45/√20),
or 10.15 < μ < 12.45.
9.13 n = 9, ¯x = 1.0056, s = 0.0245, and t0.005 = 3.355 with 8 degrees of freedom. A 99%
confidence interval for the population mean is
1.0056 − (3.355)(0.0245/3) < μ < 1.0056 + (3.355)(0.0245/3),
or 0.978 < μ < 1.033.
9.14 n = 10, ¯x = 230, s = 15, and t0.005 = 3.25 with 9 degrees of freedom. A 99% confidence
interval for the population mean is
230 − (3.25)(15/√10) < μ < 230 + (3.25)(15/√10),
or 214.58 < μ < 245.42.
9.15 n = 12, ¯x = 48.50, s = 1.5, and t0.05 = 1.796 with 11 degrees of freedom. A 90%
confidence interval for the population mean is
48.50 − (1.796)(1.5/√12) < μ < 48.50 + (1.796)(1.5/√12),
or 47.722 < μ < 49.278.
9.16 n = 12, ¯x = 79.3, s = 7.8, and t0.025 = 2.201 with 11 degrees of freedom. A 95%
confidence interval for the population mean is
79.3 − (2.201)(7.8/√12) < μ < 79.3 + (2.201)(7.8/√12),
or 74.34 < μ < 84.26.
Solutions for Exercises in Chapter 9 105
9.17 n = 25, ¯x = 325.05, s = 0.5, γ = 5%, and 1 − α = 90%, with k = 2.208. So, 325.05 ±
(2.208)(0.5) yields (323.946, 326.151). Thus, we are 95% confident that this tolerance
interval will contain 90% of the aspirin contents for this brand of buffered aspirin.
9.18 n = 15, ¯x = 3.7867, s = 0.9709, γ = 1%, and 1 − α = 95%, with k = 3.507. So, by
calculating 3.7867 ± (3.507)(0.9709) we obtain (0.382, 7.192) which is a 99% tolerance
interval that will contain 95% of the drying times.
9.19 n = 100, ¯x = 23,500, s = 3, 900, 1 − α = 0.99, and γ = 0.01, with k = 3.096. The
tolerance interval is 23,500 ± (3.096)(3,900) which yields 11,425 < μ < 35,574.
9.20 n = 12, ¯x = 48.50, s = 1.5, 1 − α = 0.90, and γ = 0.05, with k = 2.655. The tolerance
interval is 48.50 ± (2.655)(1.5) which yields (44.52, 52.48).
9.21 By definition, MSE = E(ˆ − θ)2 which can be expressed as
MSE = E[ˆ − E(ˆ ) + E(ˆ ) − θ]2
= E[ˆ − E(ˆ )]2 + E[E(ˆ ) − θ]2 + 2E[ˆ − E(ˆ )]E[E(ˆ ) − θ].
The third term on the right hand side is zero since E[ˆ − E(ˆ )] = E[ˆ ] − E(ˆ ) = 0.
Hence the claim is valid.
9.22 (a) The bias is E(S′2) − σ2 = n−1
n σ2 − σ2 = 2
n .
(b) lim
n→∞
Bias = lim
n→∞
2
n = 0.
9.23 Using Theorem 8.4, we know that X2 = (n−1)S2
2 follows a chi-squared distribution with
n − 1 degrees of freedom, whose variance is 2(n − 1). So, V ar(S2) = V ar
2
n−1X2
=
2
n−1σ4, and V ar(S′2) = V ar
n−1
n S2
=
n−1
n
2
V ar(S2) = 2(n−1) 4
n2 . Therefore, the
variance of S′2 is smaller.
9.24 Using Exercises 9.21 and 9.23,
MSE(S2)
MSE(S′2)
=
V ar(S2) + [Bias(S2)]2
V ar(S′2) + [Bias(S′2)]2 =
2σ4/(n − 1)
2(n − 1)σ4/n2 + σ4/n2
= 1 +
3n − 1
2n2 − 3n + 1
,
which is always larger than 1 when n is larger than 1. Hence the MSE of S′2 is usually
smaller.
9.25 n = 20, ¯x = 11.3, s = 2.45, and t0.025 = 2.093 with 19 degrees of freedom. A 95%
prediction interval for a future observation is
11.3 ± (2.093)(2.45)
p
1 + 1/20 = 11.3 ± 5.25,
which yields (6.05, 16.55).
106 Chapter 9 One- and Two-Sample Estimation Problems
9.26 n = 12, ¯x = 79.3, s = 7.8, and t0.025 = 2.201 with 11 degrees of freedom. A 95%
prediction interval for a future observation is
79.3 ± (2.201)(7.8)
p
1 + 1/12 = 79.3 ± 17.87,
which yields (61.43, 97.17).
9.27 n = 15, ¯x = 3.7867, s = 0.9709, and t0.025 = 2.145 with 14 degrees of freedom. A 95%
prediction interval for a new observation is
3.7867 ± (2.145)(0.9709)
p
1 + 1/15 = 3.7867 ± 2.1509,
which yields (1.6358, 5.9376).
9.28 n = 9, ¯x = 1.0056, s = 0.0245, 1 − α = 0.95, and γ = 0.05, with k = 3.532. The
tolerance interval is 1.0056 ± (3.532)(0.0245) which yields (0.919, 1.092).
9.29 n = 15, ¯x = 3.84, and s = 3.07. To calculate an upper 95% prediction limit, we
obtain t0.05 = 1.761 with 14 degrees of freedom. So, the upper limit is 3.84 +
(1.761)(3.07)
p
1 + 1/15 = 3.84 + 5.58 = 9.42. This means that a new observation
will have a chance of 95% to fall into the interval (−∞, 9.42). To obtain an upper
95% tolerance limit, using 1 − α = 0.95 and γ = 0.05, with k = 2.566, we get
3.84 + (2.566)(3.07) = 11.72. Hence, we are 95% confident that (−∞, 11.72) will
contain 95% of the orthophosphorous measurements in the river.
9.30 n = 50, ¯x = 78.3, and s = 5.6. Since t0.05 = 1.677 with 49 degrees of freedom,
the bound of a lower 95% prediction interval for a single new observation is 78.3 − (1.677)(5.6)
p
1 + 1/50 = 68.91. So, the interval is (68.91,∞). On the other hand,
with 1 − α = 95% and γ = 0.01, the k value for a one-sided tolerance limit is 2.269
and the bound is 78.3 − (2.269)(5.6) = 65.59. So, the tolerance interval is (65.59,∞).
9.31 Since the manufacturer would be more interested in the mean tensile strength for future
products, it is conceivable that prediction interval and tolerance interval may be more
interesting than just a confidence interval.
9.32 This time 1 − α = 0.99 and γ = 0.05 with k = 3.126. So, the tolerance limit is
78.3−(3.126)(5.6) = 60.79. Since 62 exceeds the lower bound of the interval, yes, this
is a cause of concern.
9.33 In Exercise 9.27, a 95% prediction interval for a new observation is calculated as
(1.6358, 5.9377). Since 6.9 is in the outside range of the prediction interval, this new
observation is likely to be an outlier.
9.34 n = 12, ¯x = 48.50, s = 1.5, 1 − α = 0.95, and γ = 0.05, with k = 2.815. The lower
bound of the one-sided tolerance interval is 48.50−(2.815)(1.5) = 44.275. Their claim
is not necessarily correct.
Solutions for Exercises in Chapter 9 107
9.35 n1 = 25, n2 = 36, ¯x1 = 80, ¯x2 = 75, σ1 = 5, σ2 = 3, and z0.03 = 1.88. So, a 94%
confidence interval for μ1 − μ2 is
(80 − 75) − (1.88)
p
25/25 + 9/36 < μ1 − μ2 < (80 − 75) + (1.88)
p
25/25 + 9/36,
which yields 2.9 < μ1 − μ2 < 7.1.
9.36 nA = 50, nB = 50, ¯xA = 78.3, ¯xB = 87.2, σA = 5.6, and σB = 6.3. It is known that
z0.025 = 1.96. So, a 95% confidence interval for the difference of the population means
is
(87.2 − 78.3) ± 1.96
p
5.62/50 + 6.32/50 = 8.9 ± 2.34,
or 6.56 < μA − μB < 11.24.
9.37 n1 = 100, n2 = 200, ¯x1 = 12.2, ¯x2 = 9.1, s1 = 1.1, and s2 = 0.9. It is known that
z0.01 = 2.327. So
(12.2 − 9.1) ± 2.327
p
1.12/100 + 0.92/200 = 3.1 ± 0.30,
or 2.80 < μ1 −μ2 < 3.40. The treatment appears to reduce the mean amount of metal
removed.
9.38 n1 = 12, n2 = 10, ¯x1 = 85, ¯x2 = 81, s1 = 4, s2 = 5, and sp = 4.478 with t0.05 = 1.725
with 20 degrees of freedom. So
(85 − 81) ± (1.725)(4.478)
p
1/12 + 1/10 = 4 ± 3.31,
which yields 0.69 < μ1 − μ2 < 7.31.
9.39 n1 = 12, n2 = 18, ¯x1 = 84, ¯x2 = 77, s1 = 4, s2 = 6, and sp = 5.305 with t0.005 = 2.763
with 28 degrees of freedom. So,
(84 − 77) ± (2.763)(5.305)
p
1/12 + 1/18 = 7 ± 5.46,
which yields 1.54 < μ1 − μ2 < 12.46.
9.40 n1 = 10, n2 = 10, ¯x1 = 0.399, ¯x2 = 0.565, s1 = 0.07279, s2 = 0.18674, and sp = 0.14172
with t0.025 = 2.101 with 18 degrees of freedom. So,
(0.565 − 0.399) ± (2.101)(0.14172)
p
1/10 + 1/10 = 0.166 ± 0.133,
which yields 0.033 < μ1 − μ2 < 0.299.
9.41 n1 = 14, n2 = 16, ¯x1 = 17, ¯x2 = 19, s2
1 = 1.5, s2
2 = 1.8, and sp = 1.289 with t0.005 =
2.763 with 28 degrees of freedom. So,
(19 − 17) ± (2.763)(1.289)
p
1/16 + 1/14 = 2 ± 1.30,
which yields 0.70 < μ1 − μ2 < 3.30.
108 Chapter 9 One- and Two-Sample Estimation Problems
9.42 n1 = 12, n2 = 10, ¯x1 = 16, ¯x2 = 11, s1 = 1.0, s2 = 0.8, and sp = 0.915 with t0.05 = 1.725
with 20 degrees of freedom. So,
(16 − 11) ± (1.725)(0.915)
p
1/12 + 1/10 = 5 ± 0.68,
which yields 4.3 < μ1 − μ2 < 5.7.
9.43 nA = nB = 12, ¯xA = 36, 300, ¯xB = 38, 100, sA = 5, 000, sB = 6, 100, and
v =
50002/12 + 61002/12
(50002/12)2
11 + (61002/12)2
11
= 21,
with t0.025 = 2.080 with 21 degrees of freedom. So,
(36, 300 − 38, 100) ± (2.080)
r
50002
12
+
61002
12
= −1, 800 ± 4, 736,
which yields −6, 536 < μA − μB < 2, 936.
9.44 n = 8, ¯ d = −1112.5, sd = 1454, with t0.005 = 3.499 with 7 degrees of freedom. So,
−1112.5 ± (3.499)
1454
√8
= −1112.5 ± 1798.7,
which yields −2911.2 < μD < 686.2.
9.45 n = 9, ¯ d = 2.778, sd = 4.5765, with t0.025 = 2.306 with 8 degrees of freedom. So,
2.778 ± (2.306)
4.5765
√9
= 2.778 ± 3.518,
which yields −0.74 < μD < 6.30.
9.46 nI = 5, nII = 7, ¯xI = 98.4, ¯xII = 110.7, sI = 8.375, and sII = 32.185, with
v =
(8.7352/5 + 32.1852/7)2
(8.7352/5)2
4 + (32.1852/7)2
6
= 7
So, t0.05 = 1.895 with 7 degrees of freedom.
(110.7 − 98.4) ± 1.895
p
8.7352/5 + 32.1852/7 = 12.3 ± 24.2,
which yields −11.9 < μII − μI < 36.5.
9.47 n = 10, ¯ d = 14.89%, and sd = 30.4868, with t0.025 = 2.262 with 9 degrees of freedom.
So,
14.89 ± (2.262)
30.4868
√10
= 14.89 ± 21.81,
which yields −6.92 < μD < 36.70.
Solutions for Exercises in Chapter 9 109
9.48 nA = nB = 20, ¯xA = 32.91, ¯xB = 30.47, sA = 1.57, sB = 1.74, and Sp = 1.657.
(a) t0.025 ≈ 2.042 with 38 degrees of freedom. So,
(32.91 − 30.47) ± (2.042)(1.657)
p
1/20 + 1/20 = 2.44 ± 1.07,
which yields 1.37 < μA − μB < 3.51.
(b) Since it is apparent that type A battery has longer life, it should be adopted.
9.49 nA = nB = 15, ¯xA = 3.82, ¯xB = 4.94, sA = 0.7794, sB = 0.7538, and sp = 0.7667 with
t0.025 = 2.048 with 28 degrees of freedom. So,
(4.94 − 3.82) ± (2.048)(0.7667)
p
1/15 + 1/15 = 1.12 ± 0.57,
which yields 0.55 < μB − μA < 1.69.
9.50 n1 = 8, n2 = 13, ¯x1 = 1.98, ¯x2 = 1.30, s1 = 0.51, s2 = 0.35, and sp = 0.416. t0.025 =
2.093 with 19 degrees of freedom. So,
(1.98 − 1.30) ± (2.093)(0.416)
p
1/8 + 1/13 = 0.68 ± 0.39,
which yields 0.29 < μ1 − μ2 < 1.07.
9.51 (a) n = 200, ˆp = 0.57, ˆq = 0.43, and z0.02 = 2.05. So,
0.57 ± (2.05)
r
(0.57)(0.43)
200
= 0.57 ± 0.072,
which yields 0.498 < p < 0.642.
(b) Error ≤ (2.05)
q
(0.57)(0.43)
200 = 0.072.
9.52 n = 500.ˆp = 485
500 = 0.97, ˆq = 0.03, and z0.05 = 1.645. So,
0.97 ± (1.645)
r
(0.97)(0.03)
500
= 0.97 ± 0.013,
which yields 0.957 < p < 0.983.
9.53 n = 1000, ˆp = 228
1000 = 0.228, ˆq = 0.772, and z0.005 = 2.575. So,
0.228 ± (2.575)
r
(0.228)(0.772)
1000
= 0.228 ± 0.034,
which yields 0.194 < p < 0.262.
9.54 n = 100, ˆp = 8
100 = 0.08, ˆq = 0.92, and z0.01 = 2.33. So,
0.08 ± (2.33)
r
(0.08)(0.92)
100
= 0.08 ± 0.063,
which yields 0.017 < p < 0.143.
110 Chapter 9 One- and Two-Sample Estimation Problems
9.55 (a) n = 40, ˆp = 34
40 = 0.85, ˆq = 0.15, and z0.025 = 1.96. So,
0.85 ± (1.96)
r
(0.85)(0.15)
40
= 0.85 ± 0.111,
which yields 0.739 < p < 0.961.
(b) Since p = 0.8 falls in the confidence interval, we can not conclude that the new
system is better.
9.56 n = 100, ˆp = 24
100 = 0.24, ˆq = 0.76, and z0.005 = 2.575.
(a) 0.24 ± (2.575)
q
(0.24)(0.76)
100 = 0.24 ± 0.110, which yields 0.130 < p < 0.350.
(b) Error ≤ (2.575)
q
(0.24)(0.76)
100 = 0.110.
9.57 n = 1600, ˆp = 2
3 , ˆq = 1
3 , and z0.025 = 1.96.
(a) 2
3 ± (1.96)
q
(2/3)(1/3)
1600 = 2
3 ± 0.023, which yields 0.644 < p < 0.690.
(b) Error ≤ (1.96)
q
(2/3)(1/3)
1600 = 0.023.
9.58 n = (1.96)2(0.32)(0.68)
(0.02)2 = 2090 when round up.
9.59 n = (2.05)2(0.57)(0.43)
(0.02)2 = 2576 when round up.
9.60 n = (2.575)2(0.228)(0.772)
(0.05)2 = 467 when round up.
9.61 n = (2.33)2(0.08)(0.92)
(0.05)2 = 160 when round up.
9.62 n = (1.96)2
(4)(0.01)2 = 9604 when round up.
9.63 n = (2.575)2
(4)(0.01)2 = 16577 when round up.
9.64 n = (1.96)2
(4)(0.04)2 = 601 when round up.
9.65 nM = nF = 1000, ˆpM = 0.250, ˆqM = 0.750, ˆpF = 0.275, ˆqF = 0.725, and z0.025 = 1.96.
So
(0.275 − 0.250) ± (1.96)
r
(0.250)(0.750)
1000
+
(0.275)(0.725)
1000
= 0.025 ± 0.039,
which yields −0.0136 < pF − pM < 0.0636.
Solutions for Exercises in Chapter 9 111
9.66 n1 = 250, n2 = 175, ˆp1 = 80
250 = 0.32, ˆp2 = 40
175 = 0.2286, and z0.05 = 1.645. So,
(0.32 − 0.2286) ± (1.645)
r
(0.32)(0.68)
250
+
(0.2286)(0.7714)
175
= 0.0914 ± 0.0713,
which yields 0.0201 < p1 − p2 < 0.1627. From this study we conclude that there is
a significantly higher proportion of women in electrical engineering than there is in
chemical engineering.
9.67 n1 = n2 = 500, ˆp1 = 120
500 = 0.24, ˆp2 = 98
500 = 0.196, and z0.05 = 1.645. So,
(0.24 − 0.196) ± (1.645)
r
(0.24)(0.76)
500
+
(0.196)(0.804)
500
= 0.044 ± 0.0429,
which yields 0.0011 < p1 − p2 < 0.0869. Since 0 is not in this confidence interval, we
conclude, at the level of 90% confidence, that inoculation has an effect on the incidence
of the disease.
9.68 n5◦C = n15◦C = 20, ˆp5◦C = 0.50, ˆp15◦C = 0.75, and z0.025 = 1.96. So,
(0.5 − 0.75) ± (1.96)
r
(0.50)(0.50)
20
+
(0.75)(0.25)
20
= −0.25 ± 0.2899,
which yields −0.5399 < p5◦C − p15◦C < 0.0399. Since this interval includes 0, the
significance of the difference cannot be shown at the confidence level of 95%.
9.69 nnow = 1000, ˆpnow = 0.2740, n91 = 760, ˆp91 = 0.3158, and z0.025 = 1.96. So,
(0.2740 − 0.3158) ± (1.96)
r
(0.2740)(0.7260)
1000
+
(0.3158)(0.6842)
760
= −0.0418 ± 0.0431,
which yields −0.0849 < pnow − p91 < 0.0013. Hence, at the confidence level of 95%,
the significance cannot be shown.
9.70 n90 = n94 = 20, ˆp90 = 0.337, and ˆ094 = 0.362
(a) n90 ˆp90 = (20)(0.337) ≈ 7 and n94 ˆp94 = (20)(0.362) ≈ 7.
(b) Since z0.025 = 1.96, (0.337−0.362)±(1.96)
q
(0.337)(0.663)
20 + (0.362)(0.638)
20 = −0.025±
0.295, which yields −0.320 < p90 − p94 < 0.270. Hence there is no evidence, at
the confidence level of 95%, that there is a change in the proportions.
9.71 s2 = 0.815 with v = 4 degrees of freedom. Also, χ2
0.025 = 11.143 and χ2
0.975 = 0.484.
So,
(4)(0.815)
11.143
< σ2 <
(4)(0.815)
0.484
, which yields 0.293 < σ2 < 6.736.
Since this interval contains 1, the claim that σ2 seems valid.
112 Chapter 9 One- and Two-Sample Estimation Problems
9.72 s2 = 16 with v = 19 degrees of freedom. It is known χ2
0.01 = 36.191 and χ2
0.99 = 7.633.
Hence
(19)(16)
36.191
< σ2 <
(19)(16)
7.633
, or 8.400 < σ2 < 39.827.
9.73 s2 = 6.0025 with v = 19 degrees of freedom. Also, χ2
0.025 = 32.852 and χ2
0.975 = 8.907.
Hence,
(19)(6.0025)
32.852
< σ2 <
(19)(6.0025)
8.907
, or 3.472 < σ2 < 12.804,
9.74 s2 = 0.0006 with v = 8 degrees of freedom. Also, χ2
0.005 = 21.955 and χ2
0.995 = 1.344.
Hence,
(8)(0.0006)
21.955
< σ2 <
(8)(0.0006)
1.344
, or 0.00022 < σ2 < 0.00357.
9.75 s2 = 225 with v = 9 degrees of freedom. Also, χ2
0.005 = 23.589 and χ2
0.995 = 1.735.
Hence,
(9)(225)
23.589
< σ2 <
(9)(225)
1.735
, or 85.845 < σ2 < 1167.147,
which yields 9.27 < σ < 34.16.
9.76 s2 = 2.25 with v = 11 degrees of freedom. Also, χ2
0.05 = 19.675 and χ2
0.95 = 4.575.
Hence,
(11)(2.25)
19.675
< σ2 <
(11)(2.25)
4.575
, or 1.258 < σ2 < 5.410.
9.77 s2
1 = 1.00, s2
2 = 0.64, f0.01(11, 9) = 5.19, and f0.01(9, 11) = 4.63. So,
1.00/0.64
5.19
<
σ2
1
σ2
2
< (1.00/0.64)(4.63), or 0.301 <
σ2
1
σ2
2
< 7.234,
which yields 0.549 < 1
2
< 2.690.
9.78 s2
1 = 50002, s2
2 = 61002, and f0.05(11, 11) = 2.82. (Note: this value can be found by
using “=finv(0.05,11,11)” in Microsoft Excel.) So,
5000
6100
2 1
2.82
<
σ2
1
σ2
2
<
5000
6100
2
(2.82), or 0.238 <
σ2
1
σ2
2
< 1.895.
Since the interval contains 1, it is reasonable to assume that σ2
1 = σ2
2.
9.79 s2
I = 76.3, s2
II = 1035.905, f0.05(4, 6) = 4.53, and f0.05(6, 4) = 6.16. So,
76.3
1035.905
1
4.53
<
σ2
I
σ2
II
<
76.3
1035.905
(6.16), or 0.016 <
σ2
I
σ2
II
< 0.454.
Hence, we may assume that σ2
I 6= σ2
II .
Solutions for Exercises in Chapter 9 113
9.80 sA = 0.7794, sB = 0.7538, and f0.025(14, 14) = 2.98 (Note: this value can be found by
using “=finv(0.025,14,14)” in Microsoft Excel.) So,
0.7794
0.7538
2
1
2.98
<
σ2
A
σ2
B
<
0.7794
0.7538
2
(2.98), or 0.623 <
σ2
A
σ2
B
< 3.186.
Hence, it is reasonable to assume the equality of the variances.
9.81 The likelihood function is
L(x1, . . . , xn) =
Yn
i=1
f(xi; p) =
Yn
i=1
pxi(1 − p)1−xi = pn¯x(1 − p)n(1−¯x).
Hence, ln L = n[¯x ln(p) + (1 − ¯x) ln(1 − p)]. Taking derivative with respect to p
and setting the derivative to zero, we obtain @ ln(L)
@p = n
¯x
p − 1−¯x
1−p
= 0, which yields
¯x
p − 1−¯x
1−p = 0. Therefore, ˆp = ¯x.
9.82 (a) The likelihood function is
L(x1, . . . , xn) =
Yn
i=1
f(xi; α, β) = (αβ)n
Yn
i=1
x −1
i e− x
i
= (αβ)ne−
n P
i
=
1
x
i
Yn
i=1
xi
! −1
.
(b) So, the log-likelihood can be expressed as
ln L = n[ln(α) + ln(β)] − α
Xn
i=1
x
i + (β − 1)
Xn
i=1
ln(xi).
To solve for the maximum likelihood estimate, we need to solve the following two
equations
∂ lnL
∂α
= 0, and
∂ ln L
∂β
= 0.
9.83 (a) The likelihood function is
L(x1, . . . , xn) =
Yn
i=1
f(xi; μ, σ) =
Yn
i=1
1
√2πσxi
e−[ln(xi)−μ]2
2 2
=
1
(2π)n/2σn
Qn
i=1
xi
exp
(
−
1
2σ2
Xn
i=1
[ln(xi) − μ]2
)
.
114 Chapter 9 One- and Two-Sample Estimation Problems
(b) It is easy to obtain
ln L = −
n
2
ln(2π) −
n
2
ln(σ2) −
Xn
i=1
ln(xi) −
1
2σ2
Xn
i=1
[ln(xi) − μ]2.
So, setting 0 = @ lnL
@μ = 1
2
Pn
i=1
[ln(xi) − μ], we obtain ˆμ = 1
n
Pn
i=1
ln(xi), and setting
0 = @ lnL
@ 2 = − n
2 2 + 1
2 4
Pn
i=1
[ln(xi) − μ]2, we get ˆσ2 = 1
n
Pn
i=1
[ln(xi) − ˆμ]2.
9.84 (a) The likelihood function is
L(x1, . . . , xn) =
1
βn (α)n
Yn
i=1
x −1
i e−xi/
=
1
βn (α)n
Yn
i=1
xi
! −1
e−
n P
i
=
1
(xi/ )
.
(b) Hence
ln L = −nα ln(β) − n ln((α)) + (α − 1)
Xn
i=1
ln(xi) −
1
β
Xn
i=1
xi.
Taking derivatives of ln L with respect to α and β, respectively and setting both
as zeros. Then solve them to obtain the maximum likelihood estimates.
9.85 L(x) = px(1 − p)1−x, and ln L = x ln(p) + (1 − x) ln(1 − p), with @ lnL
@p = x
p − 1−x
1−p = 0,
we obtain ˆp = x = 1.
9.86 From the density function b∗(x; p) =
x−1
k−1
pk(1 − p)x−k, we obtain
ln L = ln
x − 1
k − 1
+ k ln p + (n − k) ln(1 − p).
Setting @ lnL
@p = k
p − n−k
1−p = 0, we obtain ˆp = k
n.
9.87 For the estimator S2,
V ar(S2) =
1
(n − 1)2 V ar
"
Xn
i=1
(xi − ¯x)2
#
=
1
(n − 1)2V ar(σ2χ2
n−1)
=
1
(n − 1)2 σ4[2(n − 1)] =
2σ4
n − 1
.
For the estimator b σ2, we have
V ar( b σ2) =
2σ4(n − 1)
n2 .
Solutions for Exercises in Chapter 9 115
9.88 n = 7, ¯ d = 3.557, sd = 2.776, and t0.025 = 2.447 with 6 degrees of freedom. So,
3.557 ± (2.447)
2.776
√7
= 3.557 ± 2.567,
which yields 0.99 < μD < 6.12. Since 0 is not in the interval, the claim appears valid.
9.89 n = 75, x = 28, hence ˆp = 28
75 = 0.3733. Since z0.025 = 1.96, a 95% confidence interval
for p can be calculate as
0.3733 ± (1.96)
r
(0.3733)(0.6267)
75
= 0.3733 ± 0.1095,
which yields 0.2638 < p < 0.4828. Since the interval contains 0.421, the claim made
by the Roanoke Times seems reasonable.
9.90 n = 12, ¯ d = 40.58, sd = 15.791, and t0.025 = 2.201 with 11 degrees of freedom. So,
40.58 ± (2.201)
15.791
√12
= 40.58 ± 10.03,
which yields 30.55 < μD < 50.61.
9.91 n = 6, ¯ d = 1.5, sd = 1.543, and t0.025 = 2.571 with 5 degrees of freedom. So,
1.5 ± (2.571)
1.543
√6
= 1.5 ± 1.62,
which yields −0.12 < μD < 3.12.
9.92 n = 12, ¯ d = 417.5, sd = 1186.643, and t0.05 = 1.796 with 11 degrees of freedom. So,
417.5 ± (1.796)
1186.643
√12
= 417.5 ± 615.23,
which yields −197.73 < μD < 1032.73.
9.93 np = nu = 8, ¯xp = 86, 250.000, ¯xu = 79, 837.500, σp = σu = 4, 000, and z0.025 = 1.96.
So,
(86250 − 79837.5) ± (1.96)(4000)
p
1/8 + 1/8 = 6412.5 ± 3920,
which yields 2, 492.5 < μp −μu < 10, 332.5. Hence, polishing does increase the average
endurance limit.
9.94 nA = 100, nB = 120, ˆpA = 24
100 = 0.24, ˆpB = 36
120 = 0.30, and z0.025 = 1.96. So,
(0.30 − 0.24) ± (1.96)
r
(0.24)(0.76)
100
+
(0.30)(0.70)
120
= 0.06 ± 0.117,
which yields −0.057 < pB − pA < 0.177.
116 Chapter 9 One- and Two-Sample Estimation Problems
9.95 nN = nO = 23, s2
N = 105.9271, s2
O = 77.4138, and f0.025(22, 22) = 2.358. So,
105.9271
77.4138
1
2.358
<
σ2
N
σ2
O
<
105.9271
77.4138
(2.358), or 0.58 <
σ2
N
σ2
O
< 3.23.
For the ratio of the standard deviations, the 95% confidence interval is approximately
0.76 <
σN
σO
< 1.80.
Since the intervals contain 1 we will assume that the variability did not change with
the local supplier.
9.96 nA = nB = 6, ¯xA = 0.1407, ¯xB = 0.1385, sA = 0.002805, sB = 0.002665, and sp =
002736. Using a 90% confidence interval for the difference in the population means,
t0.05 = 1.812 with 10 degrees of freedom, we obtain
(0.1407 − 0.1385) ± (1.812)(0.002736)
p
1/6 + 1/6 = 0.0022 ± 0.0029,
which yields −0.0007 < μA−μB < 0.0051. Since the 90% confidence interval contains 0,
we conclude that wire A was not shown to be better than wire B, with 90% confidence.
9.97 To calculate the maximum likelihood estimator, we need to use
ln L = ln
e−nμμ
n P
i
=
1
xi
Qn
i=1
xi!
= −nμ + ln(μ)
Xn
i=1
xi − ln(
Yn
i=1
xi!).
Taking derivative with respect to μ and setting it to zero, we obtain ˆμ = 1
n
Pn
i=1
xi = ¯x.
On the other hand, using the method of moments, we also get ˆμ = ¯x.
9.98 ˆμ = ¯x and ˆσ2 = 1
n−1
Pn
i=1
(xi − ¯x)2.
9.99 Equating ¯x = eμ+ 2/2 and s2 = (e2μ+ 2 )(e 2−1), we get ln(¯x) = μ+ 2
2 , or ˆμ = ln(¯x)−ˆ 2
2 .
On the other hand, ln s2 = 2μ + σ2 + ln(e 2 − 1). Plug in the form of ˆμ, we obtain
ˆσ2 = ln
1 + s2
¯x2
.
9.100 Setting ¯x = αβ and s2 = αβ2, we get ˆα = ¯x2
s2 , and ˆ β = s2
¯x .
9.101 n1 = n2 = 300, ¯x1 = 102300, ¯x2 = 98500, s1 = 5700, and s2 = 3800.
(a) z0.005 = 2.575. Hence,
(102300 − 98500) ± (2.575)
r
57002
300
+
38002
300
= 3800 ± 1018.46,
which yields 2781.54 < μ1 − μ2 < 4818.46. There is a significant difference in
salaries between the two regions.
Solutions for Exercises in Chapter 9 117
(b) Since the sample sizes are large enough, it is not necessary to assume the normality
due to the Central Limit Theorem.
(c) We assumed that the two variances are not equal. Here we are going to obtain
a 95% confidence interval for the ratio of the two variances. It is known that
f0.025(299, 299) = 1.255. So,
5700
3800
2 1
1.255
<
σ2
1
σ2
2
<
5700
3800
2
(1.255), or 1.793 <
σ2
1
σ2
2
< 2.824.
Since the confidence interval does not contain 1, the difference between the variances
is significant.
9.102 The error in estimation, with 95% confidence, is (1.96)(4000)
q
2
n. Equating this quantity
to 1000, we obtain
n = 2
(1.96)(4000)
1000
2
= 123,
when round up. Hence, the sample sizes in Review Exercise 9.101 is sufficient to
produce a 95% confidence interval on μ1 − μ2 having a width of $1,000.
9.103 n = 300, ¯x = 6.5 and s = 2.5. Also, 1 − α = 0.99 and 1 − γ = 0.95. Using Table A.7,
k = 2.522. So, the limit of the one-sided tolerance interval is 6.5+(2.522)(2.5) = 12.805.
Since this interval contains 10, the claim by the union leaders appears valid.
9.104 n = 30, x = 8, and z0.025 = 1.96. So,
4
15 ± (1.96)
r
(4/15)(11/15)
30
=
4
15 ± 0.158,
which yields 0.108 < p < 0.425.
9.105 n = (1.96)2(4/15)(11/15)
0.052 = 301, when round up.
9.106 n1 = n2 = 100, ˆp1 = 0.1, and ˆp2 = 0.06.
(a) z0.025 = 1.96. So,
(0.1 − 0.06) ± (1.96)
r
(0.1)(0.9)
100
+
(0.06)(0.94)
100
= 0.04 ± 0.075,
which yields −0.035 < p1 − p2 < 0.115.
(b) Since the confidence interval contains 0, it does not show sufficient evidence that
p1 > p2.
118 Chapter 9 One- and Two-Sample Estimation Problems
9.107 n = 20 and s2 = 0.045. It is known that χ2
0.025 = 32.825 and χ2
0.975 = 8.907 with 19
degrees of freedom. Hence the 95% confidence interval for σ2 can be expressed as
(19)(0.045)
32.825
< σ2 <
(19)(0.045)
8.907
, or 0.012 < σ2 < 0.045.
Therefore, the 95% confidence interval for σ can be approximated as
0.110 < σ < 0.212.
Since 0.3 falls outside of the confidence interval, there is strong evidence that the
process has been improved in variability.
9.108 nA = nB = 15, ¯yA = 87, sA = 5.99, ¯yB = 75, sB = 4.85, sp = 5.450, and t0.025 = 2.048
with 28 degrees of freedom. So,
(87 − 75) ± (2.048)(5.450)
r
1
15
+
1
15
= 12 ± 4.076,
which yields 7.924 < μA − μB < 16.076. Apparently, the mean operating costs of type
A engines are higher than those of type B engines.
9.109 Since the unbiased estimators of σ2
1 and σ2
2 are S2
1 and S2
2 , respectively,
E(S2) =
1
n1 + n2 − 2
[(n1 − 1)E(S2
1 ) + (n2 − 1)E(S2
2 )] =
(n1 − 1)σ2
1 + (n2 − 1)σ2
2
n1 + n2 − 2
.
If we assume σ2
1 = σ2
2 = σ2, the right hand side of the above is σ2, which means that
S2 is unbiased for σ2.
9.110 n = 15, ¯x = 3.2, and s = 0.6.
(a) t0.01 = 2.624 with 14 degrees of freedom. So, a 99% left-sided confidence interval
has an upper bound of 3.2 + (2.624) 0.6 √15
= 3.607 seconds. We assumed normality
in the calculation.
(b) 3.2 + (2.624)(0.6)
q
1 + 1
15 = 4.826. Still, we need to assume normality in the
distribution.
(c) 1 − α = 0.99 and 1 − γ = 0.95. So, k = 3.520 with n = 15. So, the upper bound
is 3.2 + (3.520)(0.6) = 5.312. Hence, we are 99% confident to claim that 95% of
the pilot will have reaction time less than 5.312 seconds.
9.111 n = 400, x = 17, so ˆp = 17
400 = 0.0425.
(a) z0.025 = 1.96. So,
0.0425 ± (1.96)
r
(0.0425)(0.9575)
400
= 0.0425 ± 0.0198,
which yields 0.0227 < p < 0.0623.
Solutions for Exercises in Chapter 9 119
(b) z0.05 = 1.645. So, the upper bound of a left-sided 95% confidence interval is
0.0425 + (1.645)
q
(0.0425)(0.9575)
400 = 0.0591.
(c) Using both intervals, we do not have evidence to dispute suppliers’ claim.
Chapter 10
One- and Two-Sample Tests of
Hypotheses
10.1 (a) Conclude that fewer than 30% of the public are allergic to some cheese products
when, in fact, 30% or more are allergic.
(b) Conclude that at least 30% of the public are allergic to some cheese products
when, in fact, fewer than 30% are allergic.
10.2 (a) The training course is effective.
(b) The training course is effective.
10.3 (a) The firm is not guilty.
(b) The firm is guilty.
10.4 (a) α = P(X ≤ 5 | p = 0.6)+P(X ≥ 13 | p = 0.6) = 0.0338+(1−0.9729) = 0.0609.
(b) β = P(6 ≤ X ≤ 12 | p = 0.5) = 0.9963 − 0.1509 = 0.8454.
β = P(6 ≤ X ≤ 12 | p = 0.7) = 0.8732 − 0.0037 = 0.8695.
(c) This test procedure is not good for detecting differences of 0.1 in p.
10.5 (a) α = P(X < 110 | p = 0.6) + P(X > 130 | p = 0.6) = P(Z < −1.52) + P(Z >
1.52) = 2(0.0643) = 0.1286.
(b) β = P(110 < X < 130 | p = 0.5) = P(1.34 < Z < 4.31) = 0.0901.
β = P(110 < X < 130 | p = 0.7) = P(−4.71 < Z < −1.47) = 0.0708.
(c) The probability of a Type I error is somewhat high for this procedure, although
Type II errors are reduced dramatically.
10.6 (a) α = P(X ≤ 3 | p = 0.6) = 0.0548.
(b) β = P(X > 3 | p = 0.3) = 1 − 0.6496 = 0.3504.
β = P(X > 3 | p = 0.4) = 1 − 0.3823 = 0.6177.
β = P(X > 3 | p = 0.5) = 1 − 0.1719 = 0.8281.
121
122 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.7 (a) α = P(X ≤ 24 | p = 0.6) = P(Z < −1.59) = 0.0559.
(b) β = P(X > 24 | p = 0.3) = P(Z > 2.93) = 1 − 0.9983 = 0.0017.
β = P(X > 24 | p = 0.4) = P(Z > 1.30) = 1 − 0.9032 = 0.0968.
β = P(X > 24 | p = 0.5) = P(Z > −0.14) = 1 − 0.4443 = 0.5557.
10.8 (a) n = 12, p = 0.7, and α = P(X > 11) = 0.0712 + 0.0138 = 0.0850.
(b) n = 12, p = 0.9, and β = P(X ≤ 10) = 0.3410.
10.9 (a) n = 100, p = 0.7, μ = np = 70, and σ = √npq =
p
(100)(0.7)(0.3) = 4.583.
Hence z = 82.5−70
4.583 = 0.3410. Therefore,
α = P(X > 82) = P(Z > 2.73) = 1 − 0.9968 = 0.0032.
(b) n = 100, p = 0.9, μ = np = 90, and σ = √npq =
p
(100)(0.9)(0.1) = 3. Hence
z = 82.5−90
3 = −2.5. So,
β = P(X ≤ 82) = P(X < −2.5) = 0.0062.
10.10 (a) n = 7, p = 0.4, α = P(X ≤ 2) = 0.4199.
(b) n = 7, p = 0.3, β = P(X ≥ 3) = 1 − P(X ≤ 2) = 1 − 0.6471 = 0.3529.
10.11 (a) n = 70, p = 0.4, μ = np = 28, and σ = √npq = 4.099, with z = 23.5−28
4.099 = −1.10.
Then α = P(X < 24) = P(Z < −1.10) = 0.1357.
(b) n = 70, p = 0.3, μ = np = 21, and σ = √npq = 3.834, with z = 23.5−21
3.834 = 0.65
Then β = P(X ≥ 24) = P(Z > 0.65) = 0.2578.
10.12 (a) n = 400, p = 0.6, μ = np = 240, and σ = √npq = 9.798, with
z1 =
259.5 − 240
9.978
= 1.990, and z2 =
220.5 − 240
9.978
= −1.990.
Hence,
α = 2P(Z < −1.990) = (2)(0.0233) = 0.0466.
(b) When p = 0.48, then μ = 192 and σ = 9.992, with
z1 =
220.5 − 192
9.992
= 2.852, and z2 =
259.5 − 192
9.992
= 6.755.
Therefore,
β = P(2.852 < Z < 6.755) = 1 − 0.9978 = 0.0022.
10.13 From Exercise 10.12(a) we have μ = 240 and σ = 9.798. We then obtain
z1 =
214.5 − 240
9.978
= −2.60, and z2 =
265.5 − 240
9.978
= 2.60.
Solutions for Exercises in Chapter 10 123
So
α = 2P(Z < −2.60) = (2)(0.0047) = 0.0094.
Also, from Exercise 10.12(b) we have μ = 192 and σ = 9.992, with
z1 =
214.5 − 192
9.992
= 2.25, and z2 =
265.5 − 192
9.992
= 7.36.
Therefore,
β = P(2.25 < Z < 7.36) = 1 − 0.9878 = 0.0122.
10.14 (a) n = 50, μ = 15, σ = 0.5, and σ ¯X = 0.5 √50
= 0.071, with z = 14.9−15
0.071 = −1.41.
Hence, α = P(Z < −1.41) = 0.0793.
(b) If μ = 14.8, z = 14.9−14.8
0.071 = 1.41. So, β = P(Z > 1.41) = 0.0793.
If μ = 14.9, then z = 0 and β = P(Z > 0) = 0.5.
10.15 (a) μ = 200, n = 9, σ = 15 and σ ¯X = 15
3 = 5. So,
z1 =
191 − 200
5
= −1.8, and z2 =
209 − 200
5
= 1.8,
with α = 2P(Z < −1.8) = (2)(0.0359) = 0.0718.
(b) If μ = 215, then z − 1 = 191−215
5 = −4.8 and z2 = 209−215
5 = −1.2, with
β = P(−4.8 < Z < −1.2) = 0.1151 − 0 = 0.1151.
10.16 (a) When n = 15, then σ ¯X = 15
5 = 3, with μ = 200 and n = 25. Hence
z1 =
191 − 200
3
= −3, and z2 =
209 − 200
3
= 3,
with α = 2P(Z < −3) = (2)(0.0013) = 0.0026.
(b) When μ = 215, then z − 1 = 191−215
3 = −8 and z2 = 209−215
3 = −2, with
β = P(−8 < Z < −2) = 0.0228 − 0 = 0.0228.
10.17 (a) n = 50, μ = 5000, σ = 120, and σ ¯X = 120 √50
= 16.971, with z = 4970−5000
16.971 = −1.77
and α = P(Z < −1.77) = 0.0384.
(b) If μ = 4970, then z = 0 and hence β = P(Z > 0) = 0.5.
If μ = 4960, then z = 4970−4960
16.971 = 0.59 and β = P(Z > 0.59) = 0.2776.
10.18 The OC curve is shown next.
124 Chapter 10 One- and Two-Sample Tests of Hypotheses
180 190 200 210 220
0.0 0.2 0.4 0.6 0.8
OC curve
m
Probability of accepting the null hypothesis
10.19 The hypotheses are
H0 : μ = 800,
H1 : μ 6= 800.
Now, z = 788−800
40/√30
= −1.64, and P-value= 2P(Z < −1.64) = (2)(0.0505) = 0.1010.
Hence, the mean is not significantly different from 800 for α < 0.101.
10.20 The hypotheses are
H0 : μ = 5.5,
H1 : μ < 5.5.
Now, z = 5.23−5.5
0.24/√64
= −9.0, and P-value= P(Z < −9.0) ≈ 0. The White Cheddar
Popcorn, on average, weighs less than 5.5oz.
10.21 The hypotheses are
H0 : μ = 40 months,
H1 : μ < 40 months.
Now, z = 38−40
5.8/√64
= −2.76, and P-value= P(Z < −2.76) = 0.0029. Decision: reject
H0.
10.22 The hypotheses are
H0 : μ = 162.5 centimeters,
H1 : μ 6= 162.5 centimeters.
Now, z = 165.2−162.5
6.9/√50
= 2.77, and P-value= 2P(Z > 2.77) = (2)(0.0028) = 0.0056.
Decision: reject H0 and conclude that μ 6= 162.5.
Solutions for Exercises in Chapter 10 125
10.23 The hypotheses are
H0 : μ = 20, 000 kilometers,
H1 : μ > 20, 000 kilometers.
Now, z = 23,500−20,000
3900/√100
= 8.97, and P-value= P(Z > 8.97) ≈ 0. Decision: reject H0 and
conclude that μ 6= 20, 000 kilometers.
10.24 The hypotheses are
H0 : μ = 8,
H1 : μ > 8.
Now, z = 8.5−8
2.25/√225
= 3.33, and P-value= P(Z > 3.33) = 0.0004. Decision: Reject H0
and conclude that men who use TM, on average, mediate more than 8 hours per week.
10.25 The hypotheses are
H0 : μ = 10,
H1 : μ 6= 10.
α = 0.01 and df = 9.
Critical region: t < −3.25 or t > 3.25.
Computation: t = 10.06−10
0.246/√10
= 0.77.
Decision: Fail to reject H0.
10.26 The hypotheses are
H0 : μ = 220 milligrams,
H1 : μ > 220 milligrams.
α = 0.01 and df = 9.
Critical region: t > 1.729.
Computation: t = 224−220
24.5/√20
= 4.38.
Decision: Reject H0 and claim μ > 220 milligrams.
10.27 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 > μ2.
Since sp =
q
(29)(10.5)2+(29)(10.2)2
58 = 10.35, then
P
"
T >
34.0
10.35
p
1/30 + 1/30
#
= P(Z > 12.72) ≈ 0.
Hence, the conclusion is that running increases the mean RMR in older women.
126 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.28 The hypotheses are
H0 : μC = μA,
H1 : μC > μA,
with sp =
q
(24)(1.5)2+(24)(1.25)2
48 = 1.3807. We obtain t = 20.0−12.0
1.3807√2/25
= 20.48. Since
P(T > 20.48) ≈ 0, we conclude that the mean percent absorbency for the cotton fiber
is significantly higher than the mean percent absorbency for acetate.
10.29 The hypotheses are
H0 : μ = 35 minutes,
H1 : μ < 35 minutes.
α = 0.05 and df = 19.
Critical region: t < −1.729.
Computation: t = 33.1−35
4.3/√20
= −1.98.
Decision: Reject H0 and conclude that it takes less than 35 minutes, on the average,
to take the test.
10.30 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 6= μ2.
Since the variances are known, we obtain z = 81−76 √5.22/25+3.52/36
= 4.22. So, P-value≈ 0
and we conclude that μ1 > μ2.
10.31 The hypotheses are
H0 : μA − μB = 12 kilograms,
H1 : μA − μB > 12 kilograms.
α = 0.05.
Critical region: z > 1.645.
Computation: z = (86.7−77.8)−12 √(6.28)2/50+(5.61)2/50
= −2.60. So, fail to reject H0 and conclude that
the average tensile strength of thread A does not exceed the average tensile strength
of thread B by 12 kilograms.
10.32 The hypotheses are
H0 : μ1 − μ2 = $2, 000,
H1 : μ1 − μ2 > $2, 000.
Solutions for Exercises in Chapter 10 127
α = 0.01.
Critical region: z > 2.33.
Computation: z = (70750−65200)−2000 √(6000)2/200+(5000)2/200
= 6.43, with a P-value= P(Z > 6.43) ≈
0. Reject H0 and conclude that the mean salary for associate professors in research
institutions is $2000 higher than for those in other institutions.
10.33 The hypotheses are
H0 : μ1 − μ2 = 0.5 micromoles per 30 minutes,
H1 : μ1 − μ2 > 0.5 micromoles per 30 minutes.
α = 0.01.
Critical region: t > 2.485 with 25 degrees of freedom.
Computation: s2
p = (14)(1.5)2+(11)(1.2)2
25 = 1.8936, and t = (8.8−7.5)−0.5
√1.8936√1/15+1/12
= 1.50. Do
not reject H0.
10.34 The hypotheses are
H0 : μ1 − μ2 = 8,
H1 : μ1 − μ2 < 8.
Computation: s2
p = (10)(4.7)2+(16)(6.1)2
26 = 31.395, and t = (85−79)−8
√31.395√1/11+1/17
= −0.92.
Using 28 degrees of freedom and Table A.4, we obtain that 0.15 < P-value < 0.20.
Decision: Do not reject H0.
10.35 The hypotheses are
H0 : μ1 − μ2 = 0,
H1 : μ1 − μ2 < 0.
α = 0.05
Critical region: t < −1.895 with 7 degrees of freedom.
Computation: sp =
q
(3)(1.363)+(4)(3.883)
7 = 1.674, and t = 2.075−2.860
1.674√1/4+1/5
= −0.70.
Decision: Do not reject H0.
10.36 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 6= μ2.
Computation: sp =
q
51002+59002
2 = 5515, and t = 37,900−39,800
5515√1/12+1/12
= −0.84.
Using 22 degrees of freedom and since 0.20 < P(T < −0.84) < 0.3, we obtain 0.4 <
P-value < 0.6. Decision: Do not reject H0.
128 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.37 The hypotheses are
H0 : μ1 − μ2 = 4 kilometers,
H1 : μ1 − μ2 6= 4 kilometers.
α = 0.10 and the critical regions are t < −1.725 or t > 1.725 with 20 degrees of
freedom.
Computation: t = 5−4
(0.915)√1/12+1/10
= 2.55.
Decision: Reject H0.
10.38 The hypotheses are
H0 : μ1 − μ2 = 8,
H1 : μ1 − μ2 < 8.
α = 0.05 and the critical region is t < −1.714 with 23 degrees of freedom.
Computation: sp =
q
(9)(3.2)2+(14)(2.8)2
23 = 2.963, and t = 5.5−8
2.963√1/10+1/15
= −2.07.
Decision: Reject H0 and conclude that μ1 − μ2 < 8 months.
10.39 The hypotheses are
H0 : μII − μI = 10,
H1 : μII − μI > 10.
α = 0.1.
Degrees of freedom is calculated as
v =
(78.8/5 + 913.333/7)2
(78.8/5)2/4 + (913/333/7)2/6
= 7.38,
hence we use 7 degrees of freedom with the critical region t > 2.998.
Computation: t = (110−97.4)−10 √78.800/5+913.333/7
= 0.22.
Decision: Fail to reject H0.
10.40 The hypotheses are
H0 : μS = μN,
H1 : μS 6= μN.
Degrees of freedom is calculated as
v =
(0.3914782/8 + 0.2144142/24)2
(0.3914782/8)2/7 + (0.2144142/24)2/23
= 8.
Computation: t = 0.97625−0.91583 √0.3914782/8+0.2144142/24
= −0.42. Since 0.3 < P(T < −0.42) < 0.4,
we obtain 0.6 < P-value < 0.8.
Decision: Fail to reject H0.
Solutions for Exercises in Chapter 10 129
10.41 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 6= μ2.
α = 0.05.
Degrees of freedom is calculated as
v =
(7874.3292/16 + 2479/5032/12)2
(7874.3292/16)2/15 + (2479.5032/12)2/11
= 19 degrees of freedom.
Critical regions t < −2.093 or t > 2.093.
Computation: t = 9897.500−4120.833 √7874.3292/16+2479.5032/12
= 2.76.
Decision: Reject H0 and conclude that μ1 > μ2.
10.42 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 6= μ2.
α = 0.05.
Critical regions t < −2.776 or t > 2.776, with 4 degrees of freedom.
Computation: ¯ d = −0.1, sd = 0.1414, so t = −0.1
0.1414/√5
= −1.58.
Decision: Do not reject H0 and conclude that the two methods are not significantly
different.
10.43 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 > μ2.
Computation: ¯ d = 0.1417, sd = 0.198, t = 0.1417
0.198/√12
= 2.48 and 0.015 < P-value < 0.02
with 11 degrees of freedom.
Decision: Reject H0 when a significance level is above 0.02.
10.44 The hypotheses are
H0 : μ1 − μ2 = 4.5 kilograms,
H1 : μ1 − μ2 < 4.5 kilograms.
Computation: ¯ d = 3.557, sd = 2.776, t = 3.557−4.5
2.778/√7
= −0.896, and 0.2 < P-value < 0.3
with 6 degrees of freedom.
Decision: Do not reject H0.
130 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.45 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 < μ2.
Computation: ¯ d = −54.13, sd = 83.002, t = −54.13
83.002/√15
= −2.53, and 0.01 < P-value <
0.015 with 14 degrees of freedom.
Decision: Reject H0.
10.46 The hypotheses are
H0 : μ1 = μ2,
H1 : μ1 6= μ2.
α = 0.05.
Critical regions are t < −2.365 or t > 2.365 with 7 degrees of freedom.
Computation: ¯ d = 198.625, sd = 210.165, t = 198.625
210.165/√8
= 2.67.
Decision: Reject H0; length of storage influences sorbic acid residual concentrations.
10.47 n = (1.645+1.282)2(0.24)2
0.32 = 5.48. The sample size needed is 6.
10.48 β = 0.1, σ = 5.8, δ = 35.9 − 40 = −4.1. Assume α = 0.05 then z0.05 = 1.645,
z0.10 = 1.28. Therefore,
n =
(1.645 + 1.28)2(5.8)2
(−4.1)2 = 17.12 ≈ 18 due to round up.
10.49 1 − β = 0.95 so β = 0.05, δ = 3.1 and z0.01 = 2.33. Therefore,
n =
(1.645 + 2.33)2(6.9)2
3.12 = 78.28 ≈ 79 due to round up.
10.50 β = 0.05, δ = 8, α = 0.05, z0.05 = 1.645, σ1 = 6.28 and σ2 = 5.61. Therefore,
n =
(1.645 + 1.645)2(6.282 + 5.612)
82 = 11.99 ≈ 12 due to round up.
10.51 n = 1.645+0.842)2(2.25)2
[(1.2)(2.25)]2 = 4.29. The sample size would be 5.
10.52 σ = 1.25, α = 0.05, β = 0.1, δ = 0.5, so = 0.5
1.25 = 0.4. Using Table A.8 we find
n = 68.
10.53 (a) The hypotheses are
H0 : Mhot −Mcold = 0,
H1 : Mhot −Mcold 6= 0.
Solutions for Exercises in Chapter 10 131
(b) Use paired T-test and find out t = 0.99 with 0.3 < P-value < 0.4. Hence, fail to
reject H0.
10.54 Using paired T-test, we find out t = 2.4 with 8 degrees of freedom. So, 0.02 <
P-value < 0.025. Reject H0; breathing frequency significantly higher in the presence
of CO.
10.55 The hypotheses are
H0 : p = 0.40,
H1 : p > 0.40.
Denote by X for those who choose lasagna.
P-value = P(X ≥ 9 | p = 0.40) = 0.4044.
The claim that p = 0.40 is not refuted.
10.56 The hypotheses are
H0 : p = 0.40,
H1 : p > 0.40.
α = 0.05.
Test statistic: binomial variable X with p = 0.4 and n = 15.
Computation: x = 8 and np0 = (15)(0.4) = 6. Therefore, from Table A.1,
P-value = P(X ≥ 8 | p = 0.4) = 1 − P(X ≤ 7 | p = 0.4) = 0.2131,
which is larger than 0.05.
Decision: Do not reject H0.
10.57 The hypotheses are
H0 : p = 0.5,
H1 : p < 0.5.
P-value = P(X ≤ 5 | p = 0.05) = 0.0207.
Decision: Reject H0.
10.58 The hypotheses are
H0 : p = 0.6,
H1 : p < 0.6.
So
P-value ≈ P
Z <
110 − (200)(0.6) p
(200)(0.6)(0.4)
!
= P(Z < −1.44) = 0.0749.
Decision: Fail to reject H0.
132 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.59 The hypotheses are
H0 : p = 0.2,
H1 : p < 0.2.
Then
P-value ≈ P
Z <
136 − (1000)(0.2) p
(1000)(0.2)(0.8)
!
= P(Z < −5.06) ≈ 0.
Decision: Reject H0; less than 1/5 of the homes in the city are heated by oil.
10.60 The hypotheses are
H0 : p = 0.25,
H1 : p > 0.25.
α = 0.05.
Computation:
P-value ≈ P
Z >
28 − (90)(0.25) p
(90)(0.25)(0.75)
!
= P(Z > 1, 34) = 0.091.
Decision: Fail to reject H0; No sufficient evidence to conclude that p > 0.25.
10.61 The hypotheses are
H0 : p = 0.8,
H1 : p > 0.8.
α = 0.04.
Critical region: z > 1.75.
Computation: z = 250−(300)(0.8) √(300)(0.8)(0.2)
= 1.44.
Decision: Fail to reject H0; it cannot conclude that the new missile system is more
accurate.
10.62 The hypotheses are
H0 : p = 0.25,
H1 : p > 0.25.
α = 0.05.
Critical region: z > 1.645.
Computation: z = 16−(48)(0.25) √(48)(0.25)(0.75)
= 1.333.
Decision: Fail to reject H0. On the other hand, we can calculate
P-value = P(Z > 1.33) = 0.0918.
Solutions for Exercises in Chapter 10 133
10.63 The hypotheses are
H0 : p1 = p2,
H1 : p1 6= p2.
Computation: ˆp = 63+59
100+125 = 0.5422, z = (63/100)−(59/125) √(0.5422)(0.4578)(1/100+1/125)
= 2.36, with
P-value = 2P(Z > 2.36) = 0.0182.
Decision: Reject H0 at level 0.0182. The proportion of urban residents who favor the
nuclear plant is larger than the proportion of suburban residents who favor the nuclear
plant.
10.64 The hypotheses are
H0 : p1 = p2,
H1 : p1 > p2.
Computation: ˆp = 240+288
300+400 = 0.7543, z = (240/300)−(288/400) √(0.7543)(0.2457)(1/300+1/400)
= 2.44, with
P-value = P(Z > 2.44) = 0.0073.
Decision: Reject H0. The proportion of couples married less than 2 years and planning
to have children is significantly higher than that of couples married 5 years and planning
to have children.
10.65 The hypotheses are
H0 : pU = pR,
H1 : pU > pR.
Computation: ˆp = 20+10
200+150 = 0.085714, z = (20/200)−(10/150) √(0.085714)(0.914286)(1/200+1/150)
= 1.10, with
P-value = P(Z > 1.10) = 0.1357.
Decision: Fail to reject H0. It cannot be shown that breast cancer is more prevalent
in the urban community.
10.66 The hypotheses are
H0 : p1 = p2,
H1 : p1 > p2.
Computation: ˆp = 29+56
120+280 = 0.2125, z = (29/120)−(56/280)) √(0.2125)(0.7875)(1/120+1/280)
= 0.93, with
P-value = P(Z > 0.93) = 0.1762.
Decision: Fail to reject H0. There is no significant evidence to conclude that the new
medicine is more effective.
134 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.67 The hypotheses are
H0 : σ2 = 0.03,
H1 : σ2 6= 0.03.
Computation: χ2 = (9)(0.24585)2
0.03 = 18.13. Since 0.025 < P(χ2 > 18.13) < 0.05 with 9
degrees of freedom, 0.05 < P-value = 2P(χ2 > 18.13) < 0.10.
Decision: Fail to reject H0; the sample of 10 containers is not sufficient to show that
σ2 is not equal to 0.03.
10.68 The hypotheses are
H0 : σ = 6,
H1 : σ < 6.
Computation: χ2 = (19)(4.51)2
36 = 10.74. Using the table, 1−0.95 < P(χ2 < 10.74) < 0.1
with 19 degrees of freedom, we obtain 0.05 < P-value < 0.1.
Decision: Fail to reject H0; there was not sufficient evidence to conclude that the
standard deviation is less then 6 at level α = 0.05 level of significance.
10.69 The hypotheses are
H0 : σ2 = 4.2 ppm,
H1 : σ2 6= 4.2 ppm.
Computation: χ2 = (63)(4.25)2
4.2 = 63.75. Since 0.3 < P(χ2 > 63.75) < 0.5 with 63
degrees of freedom, P-value = 2P(χ2 > 18.13) > 0.6 (In Microsoft Excel, if you type
“=2*chidist(63.75,63)”, you will get the P-value as 0.8898.
Decision: Fail to reject H0; the variance of aflotoxins is not significantly different from
4.2 ppm.
10.70 The hypotheses are
H0 : σ = 1.40,
H1 : σ > 1.40.
Computation: χ2 = (11)(1.75)2
1.4 = 17.19. Using the table, 0.1 < P(χ2 > 17.19) < 0.2
with 11 degrees of freedom, we obtain 0.1 < P-value < 0.2.
Decision: Fail to reject H0; the standard deviation of the contributions from the sanitation
department is not significantly greater than $1.40 at the α = 0.01 level of
significance.
10.71 The hypotheses are
H0 : σ2 = 1.15,
H1 : σ2 > 1.15.
Solutions for Exercises in Chapter 10 135
Computation: χ2 = (24)(2.03)2
1.15 = 42.37. Since 0.01 < P(χ2 > 42.37) < 0.02 with 24
degrees of freedom, 0.01 < P-value < 0.02.
Decision: Reject H0; there is sufficient evidence to conclude, at level α = 0.05, that
the soft drink machine is out of control.
10.72 (a) The hypotheses are
H0 : σ = 10.0,
H1 : σ 6= 10.0.
Computation: z = 11.9−10.0
10.0/√200
= 2.69. So P-value = P(Z < −2.69)+P(Z > 2.69) =
0.0072. There is sufficient evidence to conclude that the standard deviation is
different from 10.0.
(b) The hypotheses are
H0 : σ2 = 6.25,
H1 : σ2 < 6.25.
Computation: z = 2.1−2.5
2.5/√144
= −1.92. P-value = P(Z < −1.92) = 0.0274.
Decision: Reject H0; the variance of the distance achieved by the diesel model is
less than the variance of the distance achieved by the gasoline model.
10.73 The hypotheses are
H0 : σ2
1 = σ2
2,
H1 : σ2
1 > σ2
2.
Computation: f = (6.1)2
(5.3)2 = 1.33. Since f0.05(10, 13) = 2.67 > 1.33, we fail to reject
H0 at level α = 0.05. So, the variability of the time to assemble the product is not
significantly greater for men. On the other hand, if you use “=fdist(1.33,10,13)”, you
will obtain the P-value = 0.3095.
10.74 The hypotheses are
H0 : σ2
1 = σ2
2,
H1 : σ2
1 6= σ2
2.
Computation: f = (7874.329)2
(2479.503)2 = 10.09. Since f0.01(15, 11) = 4.25, the P-value >
(2)(0.01) = 0.02. Hence we reject H0 at level α = 0.02 and claim that the variances
for the two locations are significantly different. The P-value = 0.0004.
10.75 The hypotheses are
H0 : σ2
1 = σ2
2,
H1 : σ2
1 6= σ2
2.
136 Chapter 10 One- and Two-Sample Tests of Hypotheses
Computation: f = 78.800
913.333 = 0.086. Since P-value = 2P(f < 0.086) = (2)(0.0164) =
0.0328 for 4 and 6 degrees of freedom, the variability of running time for company 1 is
significantly less than, at level 0.0328, the variability of running time for company 2.
10.76 The hypotheses are
H0 : σA = σB,
H1 : σA 6= σB.
Computation: f = (0.0125)
0.0108 = 1.15. Since P-value = 2P(f > 1.15) = (2)(0.424) = 0.848
for 8 and 8 degrees of freedom, the two instruments appear to have similar variability.
10.77 The hypotheses are
H0 : σ1 = σ2,
H1 : σ1 6= σ2.
Computation: f = (0.0553)2
(0.0125)2 = 19.67. Since P-value = 2P(f > 19.67) = (2)(0.0004) =
0.0008 for 7 and 7 degrees of freedom, production line 1 is not producing as consistently
as production 2.
10.78 The hypotheses are
H0 : σ1 = σ2,
H1 : σ1 6= σ2.
Computation: s1 = 291.0667 and s2 = 119.3946, f = (291.0667)2
(119.3946)2 = 5.54. Since
P-value = 2P(f > 5.54) = (2)(0.0002) = 0.0004 for 19 and 19 degrees of freedom,
hydrocarbon emissions are more consistent in the 1990 model cars.
10.79 The hypotheses are
H0 : die is balanced,
H1 : die is unbalanced.
α = 0.01.
Critical region: χ2 > 15.086 with 5 degrees of freedom.
Computation: Since ei = 30, for i = 1, 2, . . . , 6, then
χ2 =
(28 − 30)2
30
+
(36 − 30)2
30
+ · · · +
(23 − 30)2
30
= 4.47.
Decision: Fail to reject H0; the die is balanced.
Solutions for Exercises in Chapter 10 137
10.80 The hypotheses are
H0 : coin is balanced,
H1 : coin is not balanced.
α = 0.05.
Critical region: χ2 > 3.841 with 1 degrees of freedom.
Computation: Since ei = 30, for i = 1, 2, . . . , 6, then
χ2 =
(63 − 50)2
50
+
(37 − 50)2
50
= 6.76.
Decision: Reject H0; the coin is not balanced.
10.81 The hypotheses are
H0 : nuts are mixed in the ratio 5:2:2:1,
H1 : nuts are not mixed in the ratio 5:2:2:1.
α = 0.05.
Critical region: χ2 > 7.815 with 3 degrees of freedom.
Computation:
Observed 269 112 74 45
Expected 250 100 100 50
χ2 =
(269 − 250)2
250
+
(112 − 100)2
100
+
(74 − 100)2
100
+
(45 − 50)2
50
= 10.14.
Decision: Reject H0; the nuts are not mixed in the ratio 5:2:2:1.
10.82 The hypotheses are
H0 : Distribution of grades is uniform,
H1 : Distribution of grades is not uniform.
α = 0.05.
Critical region: χ2 > 9.488 with 4 degrees of freedom.
Computation: Since ei = 20, for i = 1, 2, . . . , 5, then
χ2 =
(14 − 20)2
20
+
(18 − 20)2
20
+ · · ·
(16 − 20)2
20
= 10.0.
Decision: Reject H0; the distribution of grades is not uniform.
138 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.83 The hypotheses are
H0 : Data follows the binomial distribution b(y; 3, 1/4),
H1 : Data does not follows the binomial distribution.
α = 0.01.
Computation: b(0; 3, 1/4) = 27/64, b(1; 3, 1/4) = 27/64, b(2; 3, 1/4) = 9/64, and
b(3; 3, 1/4) = 1/64. Hence e1 = 27, e2 = 27, e3 = 9 and e4 = 1. Combining the
last two classes together, we obtain
χ2 =
(21 − 27)2
27
+
(31 − 27)2
27
+
(12 − 10)2
10
= 2.33.
Critical region: χ2 > 9.210 with 2 degrees of freedom.
Decision: Fail to reject H0; the data is from a distribution not significantly different
from b(y; 3, 1/4).
10.84 The hypotheses are
H0 : Data follows the hypergeometric distribution h(x; 8, 3, 5),
H1 : Data does not follows the hypergeometric distribution.
α = 0.05.
Computation: h(0; 8, 3, 5) = 1/56, b(1; 8, 3, 5) = 15/56, b(2; 8, 3, 5) = 30/56, and
b(3; 8, 3, 5) = 10/56. Hence e1 = 2, e2 = 30, e3 = 60 and e4 = 20. Combining
the first two classes together, we obtain
χ2 =
(32 − 32)2
32
+
(55 − 60)2
60
+
(25 − 20)2
20
= 1.67.
Critical region: χ2 > 5.991 with 2 degrees of freedom.
Decision: Fail to reject H0; the data is from a distribution not significantly different
from h(y; 8, 3, 5).
10.85 The hypotheses are
H0 : f(x) = g(x; 1/2) for x = 1, 2, . . . ,
H1 : f(x) 6= g(x; 1/2).
α = 0.05.
Computation: g(x; 1/2) = 1
2x , for x = 1, 2, . . . , 7 and P(X ≥ 8) = 1
27 . Hence e1 = 128,
e2 = 64, e3 = 32, e4 = 16, e5 = 8, e6 = 4, e7 = 2 and e8 = 2. Combining the last three
classes together, we obtain
χ2 =
(136 − 128)2
128
+
(60 − 64)2
64
+
(34 − 32)2
32
+
(12 − 16)2
16
+
(9 − 8)2
8
+
(5 − 8)2
8
= 3.125
Critical region: χ2 > 11.070 with 5 degrees of freedom.
Decision: Fail to reject H0; f(x) = g(x; 1/2), for x = 1, 2, . . .
Solutions for Exercises in Chapter 10 139
10.88 The hypotheses are
H0 : Distribution of grades is normal n(x; 65, 21),
H1 : Distribution of grades is not normal.
α = 0.05.
Computation:
z values P(Z < z) P(zi−1 < Z < zi) ei oi
z1 = 19.5−65
21 = −2.17
z2 = 29.5−65
21 = −1.69
z3 = 39.5−65
21 = −1.21
0.0150
0.0454
0.1131
0.0150
0.0305
0.0676
0.9
1.8
4.1
6.8
3
2
3
8
z4 = 49.5−65
21 = −0.74
z5 = 59.5−65
21 = −0.26
z6 = 69.5−65
21 = 0.21
z7 = 79.5−65
21 = 0.69
z8 = 89.5−65
21 = 1.17
z9 = ∞
0.2296
0.3974
0.5832
0.7549
0.8790
1.0000
0.1165
0.1678
0.1858
0.1717
0.1241
0.1210
7.0
10.1
11.1
10.3
7.4
7.3
4
5
11
14
14
4
A goodness-of-fit test with 6 degrees of freedom is based on the following data:
oi 8 4 5 11 14 14 4
ei 6.8 7.0 10.1 11.1 10.3 7.4 7.3
Critical region: χ2 > 12.592.
χ2 =
(8 − 6.8)2
6.8
+
(4 − 7.0)2
7.0
+ · · · +
(4 − 7.3)2
7.3
= 12.78.
Decision: Reject H0; distribution of grades is not normal.
10.89 From the data we have
z values P(Z < z) P(zi−1 < Z < zi) ei oi
z1 = 0.795−1.8
0.4 = −2.51
z2 = 0.995−1.8
0.4 = −2.01
z3 = 1.195−1.8
0.4 = −1.51
z4 = 1.395−1.8
0.4 = −1.01
0.0060
0.0222
0.0655
0.1562
0.0060
0.0162
0.0433
0.0907
0.2
0.6
1.7
3.6
6.1
1
1
1
2
5
z5 = 1.595−1.8
0.4 = −0.51
z6 = 1.795−1.8
0.4 = −0.01
z7 = 1.995−1.8
0.4 = 0.49
z8 = 2.195−1.8
0.4 = 0.99
0.3050
0.4960
0.6879
0.8389
0.1488
0.1910
0.1919
0.1510
6.0
7.6
7.7
6.0
4
13
8
5
z9 = 2.395−1.8
0.4 = 1.49
z10 = ∞
0.9319
1.0000
0.0930
0.0681
3.7
2.7
6.4
3
2
5
140 Chapter 10 One- and Two-Sample Tests of Hypotheses
The hypotheses are
H0 : Distribution of nicotine contents is normal n(x; 1.8, 0.4),
H1 : Distribution of nicotine contents is not normal.
α = 0.01.
Computation: A goodness-of-fit test with 5 degrees of freedom is based on the following
data:
oi 5 4 13 8 5 5
ei 6.1 6.0 7.6 7.7 6.0 6.4
Critical region: χ2 > 15.086.
χ2 =
(5 − 6.1)2
6.1
+
(4 − 6.0)2
6.0
+ · · · +
(5 − 6.4)2
6.4
= 5.19.
Decision: Fail to reject H0; distribution of nicotine contents is not significantly different
from n(x; 1.8, 0.4).
10.90 The hypotheses are
H0 : Presence or absence of hypertension is independent of smoking habits,
H1 : Presence or absence of hypertension is not independent of smoking habits.
α = 0.05.
Critical region: χ2 > 5.991 with 2 degrees of freedom.
Computation:
Observed and expected frequencies
Nonsmokers Moderate Smokers Heavy Smokers Total
Hypertension 21 (33.4) 36 (30.0) 30 (23.6) 87
No Hypertension 48 (35.6) 26 (32.0) 19 (25.4) 93
Total 69 62 49 180
χ2 =
(21 − 33.4)2
33.4
+ · · · +
(19 − 25.4)2
25.4
= 14.60.
Decision: Reject H0; presence or absence of hypertension and smoking habits are not
independent.
10.91 The hypotheses are
H0 : A person’s gender and time spent watching television are independent,
H1 : A person’s gender and time spent watching television are not independent.
α = 0.01.
Critical region: χ2 > 6.635 with 1 degrees of freedom.
Computation:
Solutions for Exercises in Chapter 10 141
Observed and expected frequencies
Male Female Total
Over 25 hours 15 (20.5) 29 (23.5) 44
Under 25 hours 27 (21.5) 19 (24.5) 46
Total 42 48 90
χ2 =
(15 − 20.5)2
20.5
+
(29 − 23.5)2
23.5
+
(27 − 21.5)2
21.5
+
(19 − 24.5)2
24.5
= 5.47.
Decision: Fail to reject H0; a person’s gender and time spent watching television are
independent.
10.92 The hypotheses are
H0 : Size of family is independent of level of education of father,
H1 : Size of family and the education level of father are not independent.
α = 0.05.
Critical region: χ2 > 9.488 with 4 degrees of freedom.
Computation:
Observed and expected frequencies
Number of Children
Education 0–1 2–3 Over 3 Total
Elementary 14 (18.7) 37 (39.8) 32 (24.5) 83
Secondary 19 (17.6) 42 (37.4) 17 (23.0) 78
College 12 (8.7) 17 (18.8) 10 (11.5) 39
Total 45 96 59 200
χ2 =
(14 − 18.7)2
18.7
+
(37 − 39.8)2
39.8
+ · · · +
(10 − 11.5)2
11.5
= 7.54.
Decision: Fail to reject H0; size of family is independent of level of education of father.
10.93 The hypotheses are
H0 : Occurrence of types of crime is independent of city district,
H1 : Occurrence of types of crime is dependent upon city district.
α = 0.01.
Critical region: χ2 > 21.666 with 9 degrees of freedom.
Computation:
Observed and expected frequencies
District Assault Burglary Larceny Homicide Total
1 162 (186.4) 118 (125.8) 451 (423.5) 18 (13.3) 749
2 310 (380.0) 196 (256.6) 996 (863.4) 25 (27.1) 1527
3 258 (228.7) 193 (154.4) 458 (519.6) 10 (16.3) 919
4 280 (214.9) 175 (145.2) 390 (488.5) 19 (15.3) 864
Total 1010 682 2295 72 4059
142 Chapter 10 One- and Two-Sample Tests of Hypotheses
χ2 =
(162 − 186.4)2
186.4
+
(118 − 125.8)2
125.8
+ · · · +
(19 − 15.3)2
15.3
= 124.59.
Decision: Reject H0; occurrence of types of crime is dependent upon city district.
10.94 The hypotheses are
H0 : The three cough remedies are equally effective,
H1 : The three cough remedies are not equally effective.
α = 0.05.
Critical region: χ2 > 9.488 with 4 degrees of freedom.
Computation:
Observed and expected frequencies
NyQuil Robitussin Triaminic Total
No Relief
Some Relief
Total Relief
11 (11)
32 (29)
7 (10)
13 (11)
28 (29)
9 (10)
9 (11)
27 (29)
14 (10)
33
87
30
Total 50 50 50 150
χ2 =
(11 − 11)2
11
+
(13 − 11)2
11
+ · · · +
(14 − 10)2
10
= 3.81.
Decision: Fail to reject H0; the three cough remedies are equally effective.
10.95 The hypotheses are
H0 : The attitudes among the four counties are homogeneous,
H1 : The attitudes among the four counties are not homogeneous.
Computation:
Observed and expected frequencies
County
Attitude Craig Giles Franklin Montgomery Total
Favor 65 (74.5) 66 (55.9) 40 (37.3) 34 (37.3) 205
Oppose 42 (53.5) 30 (40.1) 33 (26.7) 42 (26.7) 147
No Opinion 93 (72.0) 54 (54.0) 27 (36.0) 24 (36.0) 198
Total 200 150 100 100 550
χ2 =
(65 − 74.5)2
74.5
+
(66 − 55.9)2
55.9
+ · · · +
(24 − 36.0)2
36.0
= 31.17.
Since P-value = P(χ2 > 31.17) < 0.001 with 6 degrees of freedom, we reject H0 and
conclude that the attitudes among the four counties are not homogeneous.
Solutions for Exercises in Chapter 10 143
10.96 The hypotheses are
H0 : The proportions of widows and widowers are equal with respect to the different
time period,
H1 : The proportions of widows and widowers are not equal with respect to the
different time period.
α = 0.05.
Critical region: χ2 > 5.991 with 2 degrees of freedom.
Computation:
Observed and expected frequencies
Years Lived Widow Widower Total
Less than 5 25 (32) 39 (32) 64
5 to 10 42 (41) 40 (41) 82
More than 10 33 (26) 21 (26) 54
Total 100 100 200
χ2 =
(25 − 32)2
32
+
(39 − 32)2
32
+ · · · +
(21 − 26)2
26
= 5.78.
Decision: Fail to reject H0; the proportions of widows and widowers are equal with
respect to the different time period.
10.97 The hypotheses are
H0 : Proportions of household within each standard of living category are equal,
H1 : Proportions of household within each standard of living category are not equal.
α = 0.05.
Critical region: χ2 > 12.592 with 6 degrees of freedom.
Computation:
Observed and expected frequencies
Period Somewhat Better Same Not as Good Total
1980: Jan.
May.
Sept.
1981: Jan.
72 (66.6)
63 (66.6)
47 (44.4)
40 (44.4)
144 (145.2)
135 (145.2)
100 (96.8)
105 (96.8)
84 (88.2)
102 (88.2)
53 (58.8)
55 (58.8)
300
300
200
200
Total 222 484 294 1000
χ2 =
(72 − 66.6)2
66.6
+
(144 − 145.2)2
145.2
+ · · · +
(55 − 58.8)2
58.8
= 5.92.
Decision: Fail to reject H0; proportions of household within each standard of living
category are equal.
144 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.98 The hypotheses are
H0 : Proportions of voters within each attitude category are the same for each of the
three states,
H1 : Proportions of voters within each attitude category are not the same for each of
the three states.
α = 0.05.
Critical region: χ2 > 9.488 with 4 degrees of freedom.
Computation:
Observed and expected frequencies
Support Do not Support Undecided Total
Indiana
Kentucky
Ohio
82 (94)
107 (94)
93 (94)
97 (79)
66 (79)
74 (79)
21 (27)
27 (27)
33 (27)
200
200
200
Total 282 237 81 600
χ2 =
(82 − 94)2
94
+
(97 − 79)2
79
+ · · · +
(33 − 27)2
27
= 12.56.
Decision: Reject H0; the proportions of voters within each attitude category are not
the same for each of the three states.
10.99 The hypotheses are
H0 : Proportions of voters favoring candidate A, candidate B, or undecided are the
same for each city,
H1 : Proportions of voters favoring candidate A, candidate B, or undecided are not
the same for each city.
α = 0.05.
Critical region: χ2 > 5.991 with 2 degrees of freedom.
Computation:
Observed and expected frequencies
Richmond Norfolk Total
Favor A
Favor B
Undecided
204 (214.5)
211 (204.5)
85 (81)
225 (214.5)
198 (204.5)
77 (81)
429
409
162
Total 500 500 1000
Solutions for Exercises in Chapter 10 145
χ2 =
(204 − 214.5)2
214.5
+
(225 − 214.5)2
214.5
+ · · · +
(77 − 81)2
81
= 1.84.
Decision: Fail to reject H0; the proportions of voters favoring candidate A, candidate
B, or undecided are not the same for each city.
10.100 The hypotheses are
H0 : p1 = p2 = p3,
H1 : p1, p2, and p3 are not all equal.
α = 0.05.
Critical region: χ2 > 5.991 with 2 degrees of freedom.
Computation:
Observed and expected frequencies
Denver Phoenix Rochester Total
Watch Soap Operas
Do not Watch
52 (48)
148 (152)
31 (36)
119 (114)
37 (36)
113 (114)
120
380
Total 200 150 150 500
χ2 =
(52 − 48)2
48
+
(31 − 36)2
36
+ · · · +
(113 − 114)2
114
= 1.39.
Decision: Fail to reject H0; no difference among the proportions.
10.101 The hypotheses are
H0 : p1 = p2,
H1 : p1 > p2.
α = 0.01.
Critical region: z > 2.33.
Computation: ˆp1 = 0.31, ˆp2 = 0.24, ˆp = 0.275, and
z =
0.31 − 0.24 p
(0.275)(0.725)(1/100 + 1/100)
= 1.11.
Decision: Fail to reject H0; proportions are the same.
10.102 Using paired t-test, we observe that t = 1.55 with P-value > 0.05. Hence, the data
was not sufficient to show that the oxygen consumptions was higher when there was
little or not CO.
10.103 (a) H0 : μ = 21.8, H1 : μ 6= 21.8; critical region in both tails.
(b) H0 : p = 0.2, H1 : p > 0.2; critical region in right tail.
146 Chapter 10 One- and Two-Sample Tests of Hypotheses
(c) H0 : μ = 6.2, H1 : μ > 6.2; critical region in right tail.
(d) H0 : p = 0.7, H1 : p < 0.7; critical region in left tail.
(e) H0 : p = 0.58, H1 : p 6= 0.58; critical region in both tails.
(f) H0 : μ = 340, H1 : μ < 340; critical region in left tail.
10.104 The hypotheses are
H0 : p1 = p2,
H1 : p1 > p2.
α = 0.05.
Critical region: z > 1.645.
Computation: ˆp1 = 0.24, ˆp2 = 0.175, ˆp = 0.203, and
z =
0.24 − 0.175 p
(0.203)(0.797)(1/300 + 1/400)
= 2.12.
Decision: Reject H0; there is statistical evidence to conclude that more Italians prefer
white champagne at weddings.
10.105 n1 = n2 = 5, ¯x1 = 165.0, s1 = 6.442, ¯x2 = 139.8, s2 = 12.617, and sp = 10.02. Hence
t =
165 − 139.8
(10.02)
p
1/5 + 1/5
= 3.98.
This is a one-sided test. Therefore, 0.0025 < P-value < 0.005 with 8 degrees of
freedom. Reject H0; the speed is increased by using the facilitation tools.
10.106 (a) H0 : p = 0.2, H1 : p > 0.2; critical region in right tail.
(b) H0 : μ = 3, H1 : μ 6= 3; critical region in both tails.
(c) H0 : p = 0.15, H1 : p < 0.15; critical region in left tail.
(d) H0 : μ = $10, H1 : μ > $10; critical region in right tail.
(e) H0 : μ = 9, H1 : μ 6= 9; critical region in both tails.
10.107 The hypotheses are
H0 : p1 = p2 = p3,
H1 : p1, p2, and p3 are not all equal.
α = 0.01.
Critical region: χ2 > 9.210 with 2 degrees of freedom.
Computation:
Solutions for Exercises in Chapter 10 147
Observed and expected frequencies
Distributor
Nuts 1 2 3 Total
Peanuts
Other
345 (339)
155 (161)
313 (339)
187 (161)
359 (339)
141 (161)
1017
483
Total 500 500 500 1500
χ2 =
(345 − 339)2
339
+
(313 − 339)2
339
+ · · · +
(141 − 161)2
161
= 10.19.
Decision: Reject H0; the proportions of peanuts for the three distributors are not equal.
10.108 The hypotheses are
H0 : p1 − p2 = 0.03,
H1 : p1 − p2 > 0.03.
Computation: ˆp1 = 0.60 and ˆp2 = 0.48.
z =
(0.60 − 0.48) − 0.03 p
(0.60)(0.40)/200 + (0.48)(0.52)/500
= 2.18.
P-value = P(Z > 2.18) = 0.0146.
Decision: Reject H0 at level higher than 0.0146; the difference in votes favoring the
proposal exceeds 3%.
10.109 The hypotheses are
H0 : p1 = p2 = p3 = p4,
H1 : p1, p2, p3, and p4 are not all equal.
α = 0.01.
Critical region: χ2 > 11.345 with 3 degrees of freedom.
Computation:
Observed and expected frequencies
Preference Maryland Virginia Georgia Alabama Total
Yes
No
65 (74)
35 (26)
71 (74)
29 (26)
78 (74)
22 (26)
82 (74)
18 (26)
296
104
Total 100 100 100 100 400
χ2 =
(65 − 74)2
74
+
(71 − 74)2
74
+ · · · +
(18 − 26)2
26
= 8.84.
Decision: Fail to reject H0; the proportions of parents favoring Bibles in elementary
schools are the same across states.
148 Chapter 10 One- and Two-Sample Tests of Hypotheses
10.110 ¯ d = −2.905, sd = 3.3557, and t = ¯ d
sd/√n = −2.12. Since 0.025 < P(T > 2.12) < 0.05
with 5 degrees of freedom, we have 0.05 < P-value < 0.10. There is no significant
change in WBC leukograms.
10.111 n1 = 15, ¯x1 = 156.33, s1 = 33.09, n2 = 18, ¯x2 = 170.00 and s2 = 30.79. First we do the
f-test to test equality of the variances. Since f = s21
s22
= 1.16 and f0.05(15, 18) = 2.27,
we conclude that the two variances are equal.
To test the difference of the means, we first calculate sp = 31.85. Therefore, t =
156.33−170.00
(31.85)√1/15+1/18
= −1.23 with a P-value > 0.10.
Decision: H0 cannot be rejected at 0.05 level of significance.
10.112 n1 = n2 = 10, ¯x1 = 7.95, s1 = 1.10, ¯x2 = 10.26 and s2 = 0.57. First we do the f-test to
test equality of the variances. Since f = s21
s22
= 3.72 and f0.05(9, 9) = 3.18, we conclude
that the two variances are not equal at level 0.10.
To test the difference of the means, we first find the degrees of freedom v = 13 when
round up. Also, t = 7.95−10.26 √1.102/10+0.572/10
= −5.90 with a P-value < 0.0005.
Decision: Reject H0; there is a significant difference in the steel rods.
10.113 n1 = n2 = 10, ¯x1 = 21.5, s1 = 5.3177, ¯x2 = 28.3 and s2 = 5.8699. Since f = s21
s22
=
0.8207 and f0.05(9, 9) = 3.18, we conclude that the two variances are equal.
sp = 5.6001 and hence t = 21.5−28.3
(5.6001)√1/10+1/10
= −2.71 with 0.005 < P-value < 0.0075.
Decision: Reject H0; the high income neighborhood produces significantly more wastewater
to be treated.
10.114 n1 = n2 = 16, ¯x1 = 48.1875, s1 = 4.9962, ¯x2 = 43.7500 and s2 = 4.6833. Since
f = s21
s22
= 1.1381 and f0.05(15, 15) = 2.40, we conclude that the two variances are equal.
sp = 4.8423 and hence t = 48.1875−43.7500
(4.8423)√1/16+1/16
= 2.59. This is a two-sided test. Since
0.005 < P(T > 2.59) < 0.0075, we have 0.01 < P-value < 0.015.
Decision: Reject H0; there is a significant difference in the number of defects.
10.115 The hypotheses are:
H0 : μ = 24 × 10−4 gm,
H1 : μ < 24 × 10−4 gm.
t = 22.8−24
4.8/√50
= −1.77 with 0.025 < P-value < 0.05. Hence, at significance level of
α = 0.05, the mean concentration of PCB in malignant breast tissue is less than
24 × 10−4 gm.
Chapter 11
Simple Linear Regression and
Correlation
11.1 (a)
P
i
xi = 778.7,
P
i
yi = 2050.0,
P
i
x2
i = 26, 591.63,
P
i
xiyi = 65, 164.04, n = 25.
Therefore,
b =
(25)(65, 164.04)− (778.7)(2050.0)
(25)(26, 591.63) − (778.7)2 = 0.5609,
a =
2050 − (0.5609)(778.7)
25
= 64.53.
(b) Using the equation ˆy = 64.53 + 0.5609x with x = 30, we find ˆy = 64.53 +
(0.5609)(30) = 81.40.
(c) Residuals appear to be random as desired.
10 20 30 40 50 60
−30 −20 −10 0 10 20 30
Arm Strength
Residual
11.2 (a)
P
i
xi = 707,
P
i
yi = 658,
P
i
x2
i = 57, 557,
P
i
xiyi = 53, 258, n = 9.
b =
(9)(53, 258) − (707)(658)
(9)(57, 557) − (707)2 = 0.7771,
a =
658 − (0.7771)(707)
9
= 12.0623.
149
150 Chapter 11 Simple Linear Regression and Correlation
Hence ˆy = 12.0623 + 0.7771x.
(b) For x = 85, ˆy = 12.0623 + (0.7771)(85) = 78.
11.3 (a)
P
i
xi = 16.5,
P
i
yi = 100.4,
P
i
x2
i = 25.85,
P
i
xiyi = 152.59, n = 11. Therefore,
b =
(11)(152.59) − (16.5)(100.4)
(11)(25.85) − (16.5)2 = 1.8091,
a =
100.4 − (1.8091)(16.5)
11
= 6.4136.
Hence ˆy = 6.4136 + 1.8091x
(b) For x = 1.75, ˆy = 6.4136 + (1.8091)(1.75) = 9.580.
(c) Residuals appear to be random as desired.
1.0 1.2 1.4 1.6 1.8 2.0
−1.0 −0.5 0.0 0.5 1.0
Temperature
Residual
11.4 (a)
P
i
xi = 311.6,
P
i
yi = 297.2,
P
i
x2
i = 8134.26,
P
i
xiyi = 7687.76, n = 12.
b =
(12)(7687.26) − (311.6)(297.2)2
= − 0.6861,
a =
297.2 − (−0.6861)(311.6)
12
= 42.582.
Hence ˆy = 42.582 − 0.6861x.
(b) At x = 24.5, ˆy = 42.582 − (0.6861)(24.5) = 25.772.
11.5 (a)
P
i
xi = 675,
P
i
yi = 488,
P
i
x2
i = 37, 125,
P
i
xiyi = 25, 005, n = 18. Therefore,
b =
(18)(25, 005) − (675)(488)
(18)(37, 125) − (675)2 = 0.5676,
a =
488 − (0.5676)(675)
18
= 5.8254.
Hence ˆy = 5.8254 + 0.5676x
Solutions for Exercises in Chapter 11 151
(b) The scatter plot and the regression line are shown below.
0 20 40 60
10 20 30 40 50
Temperature
Grams
y^ = 5.8254 + 0.5676x
(c) For x = 50, ˆy = 5.8254 + (0.5676)(50) = 34.205 grams.
11.6 (a) The scatter plot and the regression line are shown below.
40 50 60 70 80 90
20 40 60 80
Placement Test
Course Grade
y^ = 32.5059 + 0.4711x
(b)
P
i
xi = 1110,
P
i
yi = 1173,
P
i
x2
i = 67, 100,
P
i
xiyi = 67, 690, n = 20. Therefore,
b =
(20)(67, 690) − (1110)(1173)
(20)(67, 100) − (1110)2 = 0.4711,
a =
1173 − (0.4711)(1110)
20
= 32.5059.
Hence ˆy = 32.5059 + 0.4711x
(c) See part (a).
(d) For ˆy = 60, we solve 60 = 32.5059 + 0.4711x to obtain x = 58.466. Therefore,
students scoring below 59 should be denied admission.
11.7 (a) The scatter plot and the regression line are shown here.
152 Chapter 11 Simple Linear Regression and Correlation
20 25 30 35 40 45 50
400 450 500 550
Advertising Costs
Sales
y^ = 343.706 + 3.221x
(b)
P
i
xi = 410,
P
i
yi = 5445,
P
i
x2
i = 15, 650,
P
i
xiyi = 191, 325, n = 12. Therefore,
b =
(12)(191, 325)− (410)(5445)
(12)(15, 650) − (410)2 = 3.2208,
a =
5445 − (3.2208)(410)
12
= 343.7056.
Hence ˆy = 343.7056 + 3.2208x
(c) When x = $35, ˆy = 343.7056 + (3.2208)(35) = $456.43.
(d) Residuals appear to be random as desired.
20 25 30 35 40 45 50
−100 −50 0 50
Advertising Costs
Residual
11.8 (a) ˆy = −1.70 + 1.81x.
(b) ˆx = (54 + 1.71)/1.81 = 30.78.
11.9 (a)
P
i
xi = 45,
P
i
yi = 1094,
P
i
x2
i = 244.26,
P
i
xiyi = 5348.2, n = 9.
b =
(9)(5348.2) − (45)(1094)
(9)(244.26) − (45)2 = −6.3240,
a =
1094 − (−6.3240)(45)
9
= 153.1755.
Hence ˆy = 153.1755 − 6.3240x.
Solutions for Exercises in Chapter 11 153
(b) For x = 4.8, ˆy = 153.1755 − (6.3240)(4.8) = 123.
11.10 (a) ˆz = cdw, ln ˆz = ln c + (ln d)w; setting ˆy = ln z, a = ln c, b = ln d, and ˆy = a + bx,
we have
x = w 1 2 2 3 5 5
y = ln z 8.7562 8.6473 8.6570 8.5932 8.5142 8.4960
P
i
xi = 18,
P
i
yi = 51.6639,
P
i
x2
i = 68,
P
i
xiyi = 154.1954, n = 6.
b = ln d =
(6)(154.1954) − (18)(51.6639)
(6)(68) − (18)2 = −0.0569,
a = ln c =
51.6639 − (−0.0569)(18)
6
= 8.7813.
Now c = e8.7813 = 6511.3364, d = e−0.0569 = 0.9447, and ˆz = 6511.3364×0.9447w.
(b) For w = 4, ˆz = 6511.3364 × 0.94474 = $5186.16.
11.11 (a) The scatter plot and the regression line are shown here.
1300 1400 1500 1600 1700 1800
3000 3500 4000 4500 5000
Temperature
Thrust
y^ = - 1847.633 + 3.653x
(b)
P
i
xi = 14, 292,
P
i
yi = 35, 578,
P
i
x2
i = 22, 954, 054,
P
i
xiyi = 57, 441, 610, n = 9.
Therefore,
b =
(9)(57, 441, 610) − (14, 292)(35, 578)
(9)(22, 954, 054) − (14, 292)2 = 3.6529,
a =
35, 578 − (3.6529)(14, 292)
9
= −1847.69.
Hence ˆy = −1847.69 + 3.6529x.
11.12 (a) The scatter plot and the regression line are shown here.
154 Chapter 11 Simple Linear Regression and Correlation
30 40 50 60 70
250 260 270 280 290 300 310 320
Temperature
Power Consumed
y^ = 218.255 + 1.384x
(b)
P
i
xi = 401,
P
i
yi = 2301,
P
i
x2
i = 22, 495,
P
i
xiyi = 118, 652, n = 8. Therefore,
b =
(8)(118, 652) − (401)(2301))
(8)(22, 495) − (401)2 = 1.3839,
a =
2301 − (1.3839)(401)
8
= 218.26.
Hence ˆy = 218.26 + 1.3839x.
(c) For x = 65◦F, ˆy = 218.26 + (1.3839)(65) = 308.21.
11.13 (a) The scatter plot and the regression line are shown here. A simple linear model
seems suitable for the data.
30 40 50 60 70
250 260 270 280 290 300 310 320
Temperature
Power Consumed
y^ = 218.255 + 1.384x
(b)
P
i
xi = 999,
P
i
yi = 670,
P
i
x2
i = 119, 969,
P
i
xiyi = 74, 058, n = 10. Therefore,
b =
(10)(74, 058) − (999)(670)
(10)(119, 969) − (999)2 = 0.3533,
a =
670 − (0.3533)(999)
10
= 31.71.
Hence ˆy = 31.71 + 0.3533x.
(c) See (a).
Solutions for Exercises in Chapter 11 155
11.14 From the data summary, we obtain
b =
(12)(318) − [(4)(12)][(12)(12)]
(12)(232) − [(4)(12)]2 = −6.45,
a = 12 − (−6.45)(4) = 37.8.
Hence, ˆy = 37.8 − 6.45x. It appears that attending professional meetings would not
result in publishing more papers.
11.15 The least squares estimator A of α is a linear combination of normally distributed
random variables and is thus normal as well.
E(A) = E( ¯ Y − B¯x) = E( ¯ Y ) − ¯xE(B) = α + β¯x − β¯x = α,
σ2
A = σ2
¯ Y −B¯x = σ2
¯ Y + ¯x2σ2
B − 2¯xσ¯ Y B =
σ2
n
+
¯x2σ2
Pn
i=1
(xi − ¯x)2
, since σ¯ Y B = 0.
Hence
σ2
A =
Pn
i=1
x2
i
n
Pn
i=1
(xi − ¯x)2
σ2.
11.16 We have the following:
Cov( ¯ Y ,B) = E
1
n
Xn
i=1
Yi −
1
n
Xn
i=1
μYi
!
Pn
i=1
(xi − ¯x)Yi
Pn
i=1
(xi − ¯x)2 −
Pn
i=1
(xi − ¯x)μYi
Pn
i=1
(xi − ¯x)2
=
Pn
i=1
(xi − ¯x)E(Yi − μYi)2 +
P
i6=j
(xi − ¯x)E(Yi − μYi)(Yj − μYj )
n
Pn
i=1
(xi − ¯x)2
=
Pn
i=1
(xi − ¯x)σ2
Yi +
P
i6=j
Cov(Yi, Yj)
n
Pn
i=1
(xi − ¯x)2
.
Now, σ2
Yi = σ2 for all i, and Cov(Yi, Yj) = 0 for i 6= j. Therefore,
Cov( ¯ Y ,B) =
σ2
Pn
i=1
(xi − ¯x)
n
Pn
i=1
(xi − ¯x)2
= 0.
156 Chapter 11 Simple Linear Regression and Correlation
11.17 Sxx = 26, 591.63 − 778.72/25 = 2336.6824, Syy = 172, 891.46 − 20502/25 = 4791.46,
Sxy = 65, 164.04 − (778.7)(2050)/25 = 1310.64, and b = 0.5609.
(a) s2 = 4791.46−(0.5609)(1310.64)
23 = 176.362.
(b) The hypotheses are
H0 : β = 0,
H1 : β 6= 0.
α = 0.05.
Critical region: t < −2.069 or t > 2.069.
Computation: t = 0.5609 √176.362/2336.6824
= 2.04.
Decision: Do not reject H0.
11.18 Sxx = 57, 557 − 7072/9 = 2018.2222, Syy = 51, 980 − 6582/9 = 3872.8889, Sxy =
53, 258 − (707)(658)/9 = 1568.4444, a = 12.0623 and b = 0.7771.
(a) s2 = 3872.8889−(0.7771)(1568.4444)
7 = 379.150.
(b) Since s = 19.472 and t0.025 = 2.365 for 7 degrees of freedom, then a 95% confidence
interval is
12.0623 ± (2.365)
s
(379.150)(57, 557)
(9)(2018.222)
= 12.0623 ± 81.975,
which implies −69.91 < α < 94.04.
(c) 0.7771 ± (2.365)
q
379.150
2018.2222 implies −0.25 < β < 1.80.
11.19 Sxx = 25.85 − 16.52/11 = 1.1, Syy = 923.58 − 100.42/11 = 7.2018, Sxy = 152.59 −
(165)(100.4)/11 = 1.99, a = 6.4136 and b = 1.8091.
(a) s2 = 7.2018−(1.8091)(1.99)
9 = 0.40.
(b) Since s = 0.632 and t0.025 = 2.262 for 9 degrees of freedom, then a 95% confidence
interval is
6.4136 ± (2.262)(0.632)
s
25.85
(11)(1.1)
= 6.4136 ± 2.0895,
which implies 4.324 < α < 8.503.
(c) 1.8091 ± (2.262)(0.632)/√1.1 implies 0.446 < β < 3.172.
11.20 Sxx = 8134.26 − 311.62/12 = 43.0467, Syy = 7407.80 − 297.22/12 = 47.1467, Sxy =
7687.76 − (311.6)(297.2)/12 = −29.5333, a = 42.5818 and b = −0.6861.
(a) s2 = 47.1467−(−0.6861)(−29.5333)
10 = 2.688.
Solutions for Exercises in Chapter 11 157
(b) Since s = 1.640 and t0.005 = 3.169 for 10 degrees of freedom, then a 99% confidence
interval is
42.5818 ± (3.169)(1.640)
s
8134.26
(12)(43.0467)
= 42.5818 ± 20.6236,
which implies 21.958 < α < 63.205.
(c) −0.6861 ± (3.169)(1.640)/√43.0467 implies −1.478 < β < 0.106.
11.21 Sxx = 37, 125 − 6752/18 = 11, 812.5, Syy = 17, 142 − 4882/18 = 3911.7778, Sxy =
25, 005 − (675)(488)/18 = 6705, a = 5.8254 and b = 0.5676.
(a) s2 = 3911.7778−(0.5676)(6705)
16 = 6.626.
(b) Since s = 2.574 and t0.005 = 2.921 for 16 degrees of freedom, then a 99% confidence
interval is
5.8261 ± (2.921)(2.574)
s
37, 125
(18)(11, 812.5)
= 5.8261 ± 3.1417,
which implies 2.686 < α < 8.968.
(c) 0.5676 ± (2.921)(2.574)/√11, 812.5 implies 0.498 < β < 0.637.
11.22 The hypotheses are
H0 : α = 10,
H1 : α > 10.
α = 0.05.
Critical region: t > 1.734.
Computations: Sxx = 67, 100− 11102/20 = 5495, Syy = 74, 725 − 11732/20 = 5928.55,
Sxy = 67, 690 − (1110)(1173)/20 = 2588.5, s2 = 5928.55−(0.4711)(2588.5)
18 = 261.617 and
then s = 16.175. Now
t =
32.51 − 10
16.175
p
67, 100/(20)(5495)
= 1.78.
Decision: Reject H0 and claim α > 10.
11.23 The hypotheses are
H0 : β = 6,
H1 : β < 6.
α = 0.025.
Critical region: t = −2.228.
158 Chapter 11 Simple Linear Regression and Correlation
Computations: Sxx = 15, 650 − 4102/12 = 1641.667, Syy = 2, 512.925 − 54452/12 =
42, 256.25, Sxy = 191, 325 − (410)(5445)/12 = 5, 287.5, s2 = 42,256.25−(3,221)(5,287.5)
10 =
2, 522.521 and then s = 50.225. Now
t =
3.221 − 6
50.225/√1641.667
= −2.24.
Decision: Reject H0 and claim β < 6.
11.24 Using the value s = 19.472 from Exercise 11.18(a) and the fact that ¯y0 = 74.230 when
x0 = 80, and ¯x = 78.556, we have
74.230 ± (2.365)(19.472)
r
1
9
+
1.4442
2018.222
= 74.230 ± 15.4216.
Simplifying it we get 58.808 < μY | 80 < 89.652.
11.25 Using the value s = 1.64 from Exercise 11.20(a) and the fact that y0 = 25.7724 when
x0 = 24.5, and ¯x = 25.9667, we have
(a) 25.7724 ± (2.228)(1.640)
q
1
12 + (−1.4667)2
43.0467 = 25.7724 ± 1.3341 implies 24.438 <
μY | 24.5 < 27.106.
(b) 25.7724±(2.228)(1.640)
q
1 + 1
12 + (−1.4667)2
43.0467 = 25.7724±3.8898 implies 21.883 <
y0 < 29.662.
11.26 95% confidence bands are obtained by plotting the limits
(6.4136 + 1.809x) ± (2.262)(0.632)
r
1
11
+
(x − 1.5)2
1.1
.
1.0 1.2 1.4 1.6 1.8 2.0
8.0 8.5 9.0 9.5 10.0 10.5
Temperature
Converted Sugar
11.27 Using the value s = 0.632 from Exercise 11.19(a) and the fact that y0 = 9.308 when
x0 = 1.6, and ¯x = 1.5, we have
9.308 ± (2.262)(0.632)
r
1 +
1
11
+
0.12
1.1
= 9.308 ± 1.4994
implies 7.809 < y0 < 10.807.
Solutions for Exercises in Chapter 11 159
11.28 sing the value s = 2.574 from Exercise 11.21(a) and the fact that y0 = 34.205 when
x0 = 50, and ¯x = 37.5, we have
(a) 34.205±(2.921)(2.574)
q
1
18 + 12.52
11,812.5 = 34.205±1.9719 implies 32.23 < μY | 50 <
36.18.
(b) 34.205 ± (2.921)(2.574)
q
1 + 1
18 + 12.52
11,812.5 = 34.205 ± 7.7729 implies 26.43 < y0 <
41.98.
11.29 (a) 17.1812.
(b) The goal of 30 mpg is unlikely based on the confidence interval for mean mpg,
(27.95, 29.60).
(c) Based on the prediction interval, the Lexus ES300 should exceed 18 mpg.
11.30 It is easy to see that
Xn
i=1
(yi − ˆyi) =
Xn
i=1
(yi − a − bxi) =
Xn
i=1
[yi − (¯y − b¯x) − bxi)
=
Xn
i=1
(yi − ¯yi) − b
Xn
i=1
(xi − ¯x) = 0,
since a = ¯y − b¯x.
11.31 When there are only two data points x1 6= x2, using Exercise 11.30 we know that
(y1 − ˆy1) + (y2 − ˆy2) = 0. On the other hand, by the method of least squares on page
395, we also know that x1(y1 − ˆy1) + x2(y2 − ˆy2) = 0. Both of these equations yield
(x2 − x1)(y2 − ˆy2) = 0 and hence y2 − ˆy2 = 0. Therefore, y1 − ˆy1 = 0. So,
SSE = (y1 − ˆy1)2 + (y2 − ˆy2)2 = 0.
Since R2 = 1 − SSE
SST , we have R2 = 1.
11.32 (a) Suppose that the fitted model is ˆy = bx. Then
SSE =
Xn
i=1
(yi − ˆyi)2 =
Xn
i=1
(yi − bxi)2.
Taking derivative of the above with respect to b and setting the derivative to zero,
we have −2
Pn
i=1
xi(yi − bxi) = 0, which implies b =
n P
i
=
1
xiyi
n P
i
=
1
x2
i
.
(b) σ2
B =
V ar„ n
Pi=1
xiYi«
(
n P
i
=
1
x2
i )2
=
n P
i
=
1
x2
i 2
Y i
(
n P
i
=
1
x2
i )2
= 2
n P
i
=
1
x2
i
, since Yi’s are independent.
160 Chapter 11 Simple Linear Regression and Correlation
(c) E(B) =
E„ n
Pi=1
xiYi«
n P
i
=
1
x2
i
=
n P
i
=
1
xi( xi)
n P
i
=
1
x2
i
= β.
11.33 (a) The scatter plot of the data is shown next.
5 10 15 20 25 30
20 40 60 80 100
x
y
y^ = 3.416x
(b)
Pn
i=1
x2
i = 1629 and
Pn
i=1
xiyi = 5564. Hence b = 5564
1629 = 3.4156. So, ˆy = 3.4156x.
(c) See (a).
(d) Since there is only one regression coefficient, β, to be estimated, the degrees of
freedom in estimating σ2 is n − 1. So,
ˆσ2 = s2 =
SSE
n − 1
=
Pn
i=1
(yi − bxi)2
n − 1
.
(e) V ar(ˆyi) = V ar(Bxi) = x2
i V ar(B) = x2
i 2
n P
i
=
1
x2
i
.
(f) The plot is shown next.
5 10 15 20 25 30
20 40 60 80 100
x
y
11.34 Using part (e) of Exercise 11.33, we can see that the variance of a prediction y0 at x0
Solutions for Exercises in Chapter 11 161
is σ2
y0 = σ2
1 + x20
n P
i
=
1
x2
i
. Hence the 95% prediction limits are given as
(3.4145)(25) ± (2.776)√11.16132
r
1 +
252
1629
= 85.3625 ± 10.9092,
which implies 74.45 < y0 < 96.27.
11.35 (a) As shown in Exercise 11.32, the least squares estimator of β is b =
n P
i
=
1
xiyi
n P
i
=
1
x2
i
.
(b) Since
Pn
i=1
xiyi = 197.59, and
Pn
i=1
x2
i = 98.64, then b = 197.59
98.64 = 2.003 and ˆy = 2.003x.
11.36 It can be calculated that b = 1.929 and a = 0.349 and hence ˆy = 0.349 + 1.929x when
intercept is in the model. To test the hypotheses
H0 : α = 0,
H1 : α 6= 0,
with 0.10 level of significance, we have the critical regions as t < −2.132 or t > 2.132.
Computations: s2 = 0.0957 and t = 0.349 √(0.0957)(98.64)/(6)(25.14)
= 1.40.
Decision: Fail to reject H0; the intercept appears to be zero.
11.37 Now since the true model has been changed,
E(B) =
Pn
i=1
(x1i − ¯x1)E(Yi)
Pn
i=1
(x1i − ¯xi)2
=
Pn
i=1
(x1i − ¯x1)(α + βx1i + γx2i)
Pn
i=1
(x1i − ¯x1)2
=
β
Pn
i=1
(x1i − ¯x1)2 + γ
Pn
i=1
(x1i − ¯x1)x2i
Pn
i=1
(x1i − ¯x1)2
= β + γ
Pn
i=1
(x1i − ¯x1)x2i
Pn
i=1
(x1i − ¯x1)2
.
11.38 The hypotheses are
H0 : β = 0,
H1 : β 6= 0.
Level of significance: 0.05.
Critical regions: f > 5.12.
Computations: SSR = bSxy = 1.8091
1.99 = 3.60 and SSE = Syy − SSR = 7.20 − 3.60 =
3.60.
162 Chapter 11 Simple Linear Regression and Correlation
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression 3.60 1 3.60 9.00
Error 3.60 9 0.40
Total 7.20 10
Decision: Reject H0.
11.39 (a) Sxx = 1058, Syy = 198.76, Sxy = −363.63, b = Sxy
Sxx
= −0.34370, and a =
210−(−0.34370)(172.5)
25 = 10.81153.
(b) The hypotheses are
H0 : The regression is linear in x,
H1 : The regression is nonlinear in x.
α = 0.05.
Critical regions: f > 3.10 with 3 and 20 degrees of freedom.
Computations: SST = Syy = 198.76, SSR = bSxy = 124.98 and
SSE = Syy − SSR = 73.98. Since
T1. = 51.1, T2. = 51.5, T3. = 49.3, T4. = 37.0 and T5. = 22.1,
then
SSE(pure) =
X5
i=1
X5
j=1
y2
ij −
X5
i=1
T2
i.
5
= 1979.60 − 1910.272 = 69.33.
Hence the “Lack-of-fit SS” is 73.78 − 69.33 = 4.45.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
124.98
73.98
4.45
69.33
1
23
3
20
124.98
3.22
1.48
3.47
0.43
Total 198.76 24
Decision: Do not reject H0.
11.40 The hypotheses are
H0 : The regression is linear in x,
H1 : The regression is nonlinear in x.
α = 0.05.
Critical regions: f > 3.26 with 4 and 12 degrees of freedom.
Solutions for Exercises in Chapter 11 163
Computations: SST = Syy = 3911.78, SSR = bSxy = 3805.89 and SSE = Syy −
SSR = 105.89. SSE(pure) =
P6
i=1
P3
j=1
y2
ij −
P6
i=1
T2
i.
3 = 69.33, and the “Lack-of-fit SS” is
105.89 − 69.33 = 36.56.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
3805.89
105.89
36.56
69.33
1
16
4
12
3805.89
6.62
9.14
5.78
1.58
Total 3911.78 17
Decision: Do not reject H0; the lack-of-fit test is insignificant.
11.41 The hypotheses are
H0 : The regression is linear in x,
H1 : The regression is nonlinear in x.
α = 0.05.
Critical regions: f > 3.00 with 6 and 12 degrees of freedom.
Computations: SST = Syy = 5928.55, SSR = bSxy = 1219.35 and SSE = Syy −
SSR = 4709.20. SSE(pure) =
P8
i=1
Pni
j=1
y2
ij −
P8
i=1
T2
i.
ni
= 3020.67, and the “Lack-of-fit SS”
is 4709.20 − 3020.67 = 1688.53.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
1219.35
4709.20
1688.53
3020.67
1
18
6
12
1219.35
261.62
281.42
251.72
1.12
Total 5928.55 19
Decision: Do not reject H0; the lack-of-fit test is insignificant.
11.42 (a) t = 2.679 and 0.01 < P(T > 2.679) < 0.015, hence 0.02 < P-value < 0.03. There
is a strong evidence that the slope is not 0. Hence emitter drive-in time influences
gain in a positive linear fashion.
(b) f = 56.41 which results in a strong evidence that the lack-of-fit test is significant.
Hence the linear model is not adequate.
(c) Emitter does does not influence gain in a linear fashion. A better model is a
quadratic one using emitter drive-in time to explain the variability in gain.
164 Chapter 11 Simple Linear Regression and Correlation
11.43 ˆy = −21.0280 + 0.4072x; fLOF = 1.71 with a P-value = 0.2517. Hence, lack-of-fit test
is insignificant and the linear model is adequate.
11.44 (a) ˆy = 0.011571 + 0.006462x with t = 7.532 and P(T > 7.532) < 0.0005 Hence
P-value < 0.001; the slope is significantly different from 0 in the linear regression
model.
(b) fLOF = 14.02 with P-value < 0.0001. The lack-of-fit test is significant and the
linear model does not appear to be the best model.
11.45 (a) ˆy = −11.3251 − 0.0449 temperature.
(b) Yes.
(c) 0.9355.
(d) The proportion of impurities does depend on temperature.
−270 −265 −260
0.2 0.4 0.6 0.8 1.0
Temperature
Proportion of Impurity
However, based on the plot, it does not appear that the dependence is in linear
fashion. If there were replicates, a lack-of-fit test could be performed.
11.46 (a) ˆy = 125.9729 + 1.7337 population; P-value for the regression is 0.0023.
(b) f6,2 = 0.49 and P-value = 0.7912; the linear model appears to be adequate based
on the lack-of-fit test.
(c) f1,2 = 11.96 and P-value = 0.0744. The results do not change. The pure error
test is not as sensitive because the loss of error degrees of freedom.
11.47 (a) The figure is shown next.
(b) ˆy = −175.9025 + 0.0902 year; R2 = 0.3322.
(c) There is definitely a relationship between year and nitrogen oxide. It does not
appear to be linear.
Solutions for Exercises in Chapter 11 165
700 750 800 850 900 950 1000
−5 0 5 10
Time
Residual
11.48 The ANOVA model is:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
135.2000
10.4700
6.5150
3.9550
1
14
2
12
135.2000
0.7479
3.2575
0.3296
9.88
Total 145.6700 15
The P-value = 0.0029 with f = 9.88.
Decision: Reject H0; the lack-of-fit test is significant.
11.49 Sxx = 36, 354 − 35, 882.667 = 471.333, Syy = 38, 254 − 37, 762.667 = 491.333, and
Sxy = 36, 926 − 36, 810.667 = 115.333. So, r = 115 √(471.333)(491.333)
= 0.240.
11.50 The hypotheses are
H0 : ρ = 0,
H1 : ρ 6= 0.
α = 0.05.
Critical regions: t < −2.776 or t > 2.776.
Computations: t = 0.240√4 √1−0.2402 = 0.51.
Decision: Do not reject H0.
11.51 Since b = Sxy
Sxx
, we can write s2 = Syy−bSxy
n−2 = Syy−b2Sxx
n−2 . Also, b = r
q
Syy
Sxx
so that
s2 = Syy−r2Syy
n−2 = (1−r2)Syy
n−2 , and hence
t =
b
s√Sxx
=
r
p
p Syy/Sxx
SyySxx(1 − r2)/(n − 2)
=
r√n − 2
√1 − r2
.
166 Chapter 11 Simple Linear Regression and Correlation
11.52 (a) Sxx = 128.6602 − 32.682/9 = 9.9955, Syy = 7980.83 − 266.72/9 = 77.62, and
Sxy = 990.268 − (32.68)(266.7)/9 = 21.8507. So, r = 21.8507 √(9.9955)(77.62)
= 0.784.
(b) The hypotheses are
H0 : ρ = 0,
H1 : ρ > 0.
α = 0.01.
Critical regions: t > 2.998.
Computations: t = 0.784√7 √1−0.7842 = 3.34.
Decision: Reject H0; ρ > 0.
(c) (0.784)2(100%) = 61.5%.
11.53 (a) From the data of Exercise 11.1 we can calculate
Sxx = 26, 591.63 − (778.7)2/25 = 2336.6824,
Syy = 172, 891.46 − (2050)2/25 = 4791.46,
Sxy = 65, 164.04 − (778.7)(2050)/25 = 1310.64.
Therefore, r = 1310.64 √(2236.6824)(4791.46)
= 0.392.
(b) The hypotheses are
H0 : ρ = 0,
H1 : ρ 6= 0.
α = 0.05.
Critical regions: t < −2.069 or t > 2.069.
Computations: t = 0.392√23 √1−0.3922 = 2.04.
Decision: Fail to reject H0 at level 0.05. However, the P-value = 0.0530 which is
marginal.
11.54 (a) From the data of Exercise 11.9 we find Sxx = 244.26 − 452/9 = 19.26, Syy =
133, 786 − 10942/9 = 804.2222, and Sxy = 5348.2 − (45)(1094)/9 = −121.8. So,
r = −121.8 √(19.26)(804.2222)
= −0.979.
(b) The hypotheses are
H0 : ρ = −0.5,
H1 : ρ < −0.5.
α = 0.025.
Critical regions: z < −1.96.
Computations: z =
√6
2 ln
h
(0.021)(1.5)
(1.979)(0.5)
i
= −4.22.
Decision: Reject H0; ρ < −0.5.
Solutions for Exercises in Chapter 11 167
(c) (−0.979)2(100%) = 95.8%.
11.55 Using the value s = 16.175 from Exercise 11.6 and the fact that ˆy0 = 48.994 when
x0 = 35, and ¯x = 55.5, we have
(a) 48.994±(2.101)(16.175)
p
1/20 + (−20.5)2/5495 which implies to 36.908 < μY | 35 <
61.080.
(b) 48.994 ± (2.101)(16.175)
p
1 + 1/20 + (−20.5)2/5495 which implies to 12.925 <
y0 < 85.063.
11.56 The fitted model can be derived as ˆy = 3667.3968 − 47.3289x.
The hypotheses are
H0 : β = 0,
H1 : β 6= 0.
t = −0.30 with P-value = 0.77. Hence H0 cannot be rejected.
11.57 (a) Sxx = 729.18 − 118.62/20 = 25.882, Sxy = 1714.62 − (118.6)(281.1)/20 = 47.697,
so b = Sxy
Sxx
= 1.8429, and a = ¯y − b¯x = 3.1266. Hence ˆy = 3.1266 + 1.8429x.
(b) The hypotheses are
H0 : the regression is linear in x,
H1 : the regression is not linear in x.
α = 0.05.
Critical region: f > 3.07 with 8 and 10 degrees of freedom.
Computations: SST = 13.3695, SSR = 87.9008, SSE = 50.4687, SSE(pure) =
16.375, and Lack-of-fit SS = 34.0937.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
87.9008
50.4687
34.0937
16.375
1
18
8
10
87.9008
2.8038
4.2617
1.6375
2.60
Total 138.3695 19
The P-value = 0.0791. The linear model is adequate at the level 0.05.
11.58 Using the value s = 50.225 and the fact that ˆy0 = $448.644 when x0 = $45, and
¯x = $34.167, we have
(a) 488.644 ± (1.812)(50.225)
q
1/12 + 10.8332
1641.667 , which implies 452.835 < μY | 45 <
524.453.
168 Chapter 11 Simple Linear Regression and Correlation
(b) 488.644 ± (1.812)(50.225)
q
1 + 1/12 + 10.8332
1641.667 , which implies 390.845 < y0 <
586.443.
11.59 (a) ˆy = 7.3598 + 135.4034x.
(b) SS(Pure Error) = 52, 941.06; fLOF = 0.46 with P-value = 0.64. The lack-of-fit
test is insignificant.
(c) No.
11.60 (a) Sxx = 672.9167, Syy = 728.25, Sxy = 603.75 and r = 603.75 √(672.9167)(728.25)
= 0.862,
which means that (0.862)2(100%) = 74.3% of the total variation of the values of
Y in our sample is accounted for by a linear relationship with the values of X.
(b) To estimate and test hypotheses on ρ, X and Y are assumed to be random
variables from a bivariate normal distribution.
(c) The hypotheses are
H0 : ρ = 0.5,
H1 : ρ > 0.5.
α = 0.01.
Critical regions: z > 2.33.
Computations: z =
√9
2 ln
h
(1.862)(0.5)
(0.138)(1.5)
i
= 2.26.
Decision: Reject H0; ρ > 0.5.
11.61 s2 =
n P
i
=
1
(yi−ˆyi)2
n−2 . Using the centered model, ˆyi = ¯y + b(xi − ¯x) + ǫi.
(n − 2)E(S2) = E
Xn
i=1
[α + β(xi − ¯x) + ǫi − (¯y + b(xi − ¯x))]2
=
Xn
i=1
E
(α − ¯y)2 + (β − b)2(xi − ¯x)2 + ǫ2
i − 2b(xi − ¯x)ǫi − 2¯yǫi
,
(other cross product terms go to 0)
=
nσ2
n
+
σ2Sxx
Sxx
+ nσ2 − 2
σ2Sxx
Sxx − 2
nσ2
n
= (n − 2)σ2.
11.62 (a) The confidence interval is an interval on the mean sale price for a given buyer’s
bid. The prediction interval is an interval on a future observed sale price for a
given buyer’s bid.
(b) The standard errors of the prediction of sale price depend on the value of the
buyer’s bid.
Solutions for Exercises in Chapter 11 169
(c) Observations 4, 9, 10, and 17 have the lowest standard errors of prediction. These
observations have buyer’s bids very close to the mean.
11.63 (a) The residual plot appears to have a pattern and not random scatter. The R2 is
only 0.82.
(b) The log model has an R2 of 0.84. There is still a pattern in the residuals.
(c) The model using gallons per 100 miles has the best R2 with a 0.85. The residuals
appear to be more random. This model is the best of the three models attempted.
Perhaps a better model could be found.
11.64 (a) The plot of the data and an added least squares fitted line are given here.
150 200 250 300
75 80 85 90 95
Temperature
Yield
(b) Yes.
(c) ˆy = 61.5133 + 0.1139x.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Regression
E rror
Lack of fit
Pure error
486.21
24.80
3.61
21.19
1
10
2
8
486.21
2.48
1.81
2.65
0.68
Total 511.01 11
The P-value = 0.533.
(d) The results in (c) show that the linear model is adequate.
11.65 (a) ˆy = 90.8904 − 0.0513x.
(b) The t-value in testing H0 : β = 0 is −6.533 which results in a P-value < 0.0001.
Hence, the time it takes to run two miles has a significant influence on maximum
oxygen uptake.
(c) The residual graph shows that there may be some systematic behavior of the
residuals and hence the residuals are not completely random
170 Chapter 11 Simple Linear Regression and Correlation
700 750 800 850 900 950 1000
−5 0 5 10
Time
Residual
11.66 Let Y ∗ i = Yi − α, for i = 1, 2, . . . , n. The model Yi = α + βxi + ǫi is equivalent to
Y ∗ i = βxi + ǫi. This is a “regression through the origin” model that is studied in
Exercise 11.32.
(a) Using the result from Exercise 11.32(a), we have
b =
Pn
i=1
xi(yi − α)
Pn
i=1
x2
i
=
Pn
i=1
xiyi − n¯xα
Pn
i=1
x2
i
.
(b) Also from Exercise 11.32(b) we have σ2
B = 2
n P
i
=
1
x2
i
.
11.67 SSE =
Pn
i=1
(yi − βxi)2. Taking derivative with respect to β and setting this as 0, we
get
Pn
i=1
xi(yi − bxi) = 0, or
Pn
i=1
xi(yi − ˆyi) = 0. This is the only equation we can get
using the least squares method. Hence in general,
Pn
i=1
(yi − ˆyi) = 0 does not hold for a
regression model with zero intercept.
11.68 No solution is provided.
Chapter 12
Multiple Linear Regression and
Certain Nonlinear Regression Models
12.1 (a) by = 27.5467 + 0.9217x1 + 0.2842x2.
(b) When x1 = 60 and x2 = 4, the predicted value of the chemistry grade is
ˆy = 27.5467 + (0.9217)(60) + (0.2842)(4) = 84.
12.2 ˆy = −3.3727 + 0.0036x1 + 0.9476x2.
12.3 ˆy = 0.7800 + 2.7122x1 + 2.0497x2.
12.4 (a) ˆy = −22.99316 + 1.39567x1 + 0.21761x2.
(b) ˆy = −22.99316 + (1.39567)(35) + (0.21761)(250) = 80.25874.
12.5 (a) ˆy = 56.46333 + 0.15253x − 0.00008x2.
(b) ˆy = 56.46333 + (0.15253)(225)− (0.00008)(225)2 = 86.73333%.
12.6 (a) ˆ d = 13.35875 − 0.33944v − 0.01183v2.
(b) ˆ d = 13.35875 − (−0.33944)(70) − (0.01183)(70)2 = 47.54206.
12.7 ˆy = 141.61178 − 0.28193x + 0.00031x2.
12.8 (a) ˆy = 19.03333 + 1.0086x − 0.02038x2.
(b) SSE = 24.47619 with 12 degrees of freedom and SS(pure error) = 24.36667
with 10 degrees of freedom. So, SSLOF = 24.47619 − 24.36667 = 0.10952 with
2 degrees of freedom. Hence f = 0.10952/2
24.36667/10 = 0.02 with a P-value of 0.9778.
Therefore, there is no lack of fit and the quadratic model fits the data well.
12.9 (a) ˆy = −102.71324 + 0.60537x1 + 8.92364x2 + 1.43746x3 + 0.01361x4.
(b) ˆy = −102.71324+(0.60537)(75)+(8.92364)(24)+(1.43746)(90)+(0.01361)(98) =
287.56183.
171
172 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
12.10 (a) ˆy = 1.07143 + 4.60317x − 1.84524x2 + 0.19444x3.
(b) ˆy = 1.07143 + (4.60317)(2) − (1.84524)(2)2 + (0.19444)(2)3 = 4.45238.
12.11 ˆy = 3.3205 + 0.42105x1 − 0.29578x2 + 0.01638x3 + 0.12465x4.
12.12 ˆy = 1, 962.94816 − 15.85168x1 + 0.05593x2 + 1.58962x3 − 4.21867x4 − 394.31412x5.
12.13 ˆy = −6.51221 + 1.99941x1 − 3.67510x2 + 2.52449x3 + 5.15808x4 + 14.40116x5.
12.14 ˆy = −884.667 − 3 − 0.83813x1 + 4.90661x2 + 1.33113x3 + 11.93129x4.
12.15 (a) ˆy = 350.99427 − 1.27199x1 − 0.15390x2.
(b) ˆy = 350.99427 − (1.27199)(20) − (0.15390)(1200) = 140.86930.
12.16 (a) ˆy = −21.46964 − 3.32434x1 + 0.24649x2 + 20.34481x3.
(b) ˆy = −21.46964 − (3.32434)(14) + (0.24649)(220) + (20.34481)(5) = 87.94123.
12.17 s2 = 0.16508.
12.18 s2 = 0.43161.
12.19 s2 = 242.71561.
12.20 Using SAS output, we obtain
ˆσ2
b1 = 3.747 × 10−7, ˆσ2
b2 = 0.13024, ˆσb1b2 = −4.165 × 10−7.
12.21 Using SAS output, we obtain
(a) ˆσ2
b2 = 28.09554.
(b) ˆσb1b4 = −0.00958.
12.22 Using SAS output, we obtain
0.4516 < μY |x1=900,x2=1 < 1.2083, and −0.1640 < y0 < 1.8239.
12.23 Using SAS output, we obtain a 90% confidence interval for the mean response when
x = 19.5 as 29.9284 < μY |x=19.5 < 31.9729.
12.24 Using SAS output, we obtain
263.7879 < μY |x1=75,x2=24,x3=90,x4=98 < 311.3357, and 243.7175 < y0 < 331.4062.
12.25 The hypotheses are
H0 : β2 = 0,
H1 : β2 6= 0.
The test statistic value is t = 2.86 with a P-value = 0.0145. Hence, we reject H0 and
conclude β2 6= 0.
Solutions for Exercises in Chapter 12 173
12.26 The test statistic is t = 0.00362
0.000612 = 5.91 with P-value = 0.0002. Reject H0 and claim
that β1 6= 0.
12.27 The hypotheses are
H0 : β1 = 2,
H1 : β1 6= 2.
The test statistics is t = 2.71224−2
0.20209 = 3.524 with P-value = 0.0097. Reject H0 and
conclude that β1 6= 2.
12.28 Using SAS output, we obtain
(a) s2 = 650.1408.
(b) ˆy = 171.6501, 135.8735 < μY |x1=20,x2=1000 < 207.4268, and
82.9677 < y0 < 260.3326.
12.29 (a) P-value = 0.3562. Hence, fail to reject H0.
(b) P-value = 0.1841. Again, fail to reject H0.
(c) There is not sufficient evidence that the regressors x1 and x2 significantly influence
the response with the described linear model.
12.30 (a) s2 = 17.22858.
(b) ˆy = 104.9617 and 95.5660 < y0 < 114.3574.
12.31 R2 = SSR
SST = 10953
10956 = 99.97%. Hence, 99.97% of the variation in the response Y in our
sample can be explained by the linear model.
12.32 The hypotheses are:
H0 : β1 = β2 = 0,
H1 : At least one of the βi’s is not zero, for i = 1, 2.
Using the f-test, we obtain that f = MSR
MSE = 5476.60129
0.43161 = 12688.7 with P-value <
0.0001. Hence, we reject H0. The regression explained by the model is significant.
12.33 f = 5.11 with P-value = 0.0303. At level of 0.01, we fail to reject H0 and we cannot
claim that the regression is significant.
12.34 The hypotheses are:
H0 : β1 = β2 = 0,
H1 : At least one of the βi’s is not zero, for i = 1, 2.
174 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
The partial f-test statistic is
f =
(160.93598 − 145.88354)/2
1.49331
= 5.04, with 2 and 7 degrees of freedom.
The resulting P-value = 0.0441. Therefore, we reject H0 and claim that at least one
of β1 and β2 is not zero.
12.35 f = (6.90079−1.13811)/1
0.16508 = 34.91 with 1 and 9 degrees of freedom. The P-value = 0.0002
which implies that H0 is rejected.
12.36 (a) ˆy = 0.900 + 0.575x1 + 0.550x2 + 1.150x3.
(b) For the model in (a), SSR = 15.645, SSE = 1.375 and SST = 17.020. The
ANOVA table for all these single-degree-of-freedom components can be displayed
as:
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
x1
x2
x3
Error
1
1
1
4
2.645
2.420
10.580
0.34375
7.69
7.04
30.78
0.0501
0.0568
0.0052
Total 7
β3 is found to be significant at the 0.01 level and β1 and β2 are not significant.
12.37 The hypotheses are:
H0 : β1 = β2 = 0,
H1 : At least one of the βi’s is not zero, for i = 1, 2.
The partial f-test statistic is
f =
(4957.24074 − 17.02338)/2
242.71561
= 10.18, with 2 and 7 degrees of freedom.
The resulting P-value = 0.0085. Therefore, we reject H0 and claim that at least one
of β1 and β2 is not zero.
12.38 Using computer software, we obtain the following.
R(β1 | β0) = 2.645,
R(β1 | β0, β2, β3) = R(β0, β1, β2, β3) − R(β0, β2, β3) = 15.645 − 13.000 = 2.645.
R(β2 | β0, β1) = R(β0, β1, β2) − R(β0, β1) = 5.065 − 2.645 = 2.420,
R(β2 | β0, β1, β3) = R(β0, β1, β2, β3) − R(β0, β1, β3) = 15.645 − 13.225 = 2.420,
R(β3 | β0, β1, β2) = R(β0, β1, β2, β3) − R(β0, β1, β2) = 15.645 − 5.065 = 10.580.
Solutions for Exercises in Chapter 12 175
12.39 The following is the summary.
s2 R2 R2
adj
The model using weight alone 8.12709 0.8155 0.8104
The model using weight and drive ratio 4.78022 0.8945 0.8885
The above favor the model using both explanatory variables. Furthermore, in the
model with two independent variables, the t-test for β2, the coefficient of drive ratio,
shows P-value < 0.0001. Hence, the drive ratio variable is important.
12.40 The following is the summary:
s2 C.V. R2
adj Average Length of the CIs
The model with x3 4.29738 7.13885 0.8823 5.03528
The model without x3 4.00063 6.88796 0.8904 4.11769
These numbers favor the model without using x3. Hence, variable x3 appears to be
unimportant.
12.41 The following is the summary:
s2 C.V. R2
adj
The model with 3 terms 0.41966 4.62867 0.9807
The model without 3 terms 1.60019 9.03847 0.9266
Furthermore, to test β11 = β12 = β22 = 0 using the full model, f = 15.07 with
P-value = 0.0002. Hence, the model with interaction and pure quadratic terms is
better.
12.42 (a) Full model: ˆy = 121.75 + 2.50x1 + 14.75x2 + 21.75x3, with R2
adj = 0.9714.
Reduced model: ˆy = 121.75 + 14.75x2 + 21.75x3, with R2
adj = 0.9648.
There appears to be little advantage using the full model.
(b) The average prediction interval widths are:
full model: 32.70; and reduced model: 32.18. Hence, the model without using x1
is very competitive.
12.43 The following is the summary:
s2 C.V. R2
adj Average Length of the CIs
x1, x2 650.14075 16.55705 0.7696 106.60577
x1 967.90773 20.20209 0.6571 94.31092
x2 679.99655 16.93295 0.7591 78.81977
In addition, in the full model when the individual coefficients are tested, we obtain
P-value = 0.3562 for testing β1 = 0 and P-value = 0.1841 for testing β2 = 0.
In comparing the three models, it appears that the model with x2 only is slightly
better.
176 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
12.44 Here is the summary for all four models (including the full model)
s2 C.V. R2
adj
x1, x2, x3 17.22858 3.78513 0.9899
x1, x2 297.97747 15.74156 0.8250
x1, x3 17.01876 3.76201 0.9900
x2, x3 17.07575 3.76830 0.9900
It appears that a two-variable model is very competitive with the full model as long as
the model contains x3.
12.45 (a) ˆy = 5.95931−0.00003773 odometer +0.33735 octane −12.62656 van −12.98455 suv.
(b) Since the coefficients of van and suv are both negative, sedan should have the best
gas mileage.
(c) The parameter estimates (standard errors) for van and suv are −12.63 (1.07) and
−12, 98 (1.11), respectively. So, the difference between the estimates are smaller
than one standard error of each. So, no significant difference in a van and an suv
in terms of gas mileage performance.
12.46 The parameter estimates are given here.
Variable DF Estimate Standar Error t P-value
Intercept
Income
Family
Female
1
1
1
1
−206.64625
0.00543
−49.24044
236.72576
163.70943
0.00274
51.95786
110.57158
−1.26
1.98
−0.95
2.14
0.2249
0.0649
0.3574
0.0480
(a) ˆy = −206.64625 + 0.00543Income − 49.24044Family + 236/72576Female. The
company would prefer female customers.
(b) Since the P-value = 0.0649 for the coefficient of the “Income,” it is at least
marginally important. Note that the R2 = 0.3075 which is not very high. Perhaps
other variables need to be considered.
12.47 (a) Han\g Time = 1.10765 + 0.01370 LLS + 0.00429 Power.
(b) Han\g Time = 1.10765 + (0.01370)(180) + (0.00429)(260) = 4.6900.
(c) 4.4502 < μHang Time | LLS=180, Power=260 < 4.9299.
12.48 (a) For forward selection, variable x1 is entered first, and no other variables are entered
at 0.05 level. Hence the final model is ˆy = −6.33592 + 0.33738x1.
(b) For the backward elimination, variable x3 is eliminated first, then variable x4 and
then variable x2, all at 0.05 level of significance. Hence only x1 remains in the
model and the final model is the same one as in (a).
Solutions for Exercises in Chapter 12 177
(c) For the stepwise regression, after x1 is entered, no other variables are entered.
Hence the final model is still the same one as in (a) and (b).
12.49 Using computer output, with α = 0.05, x4 was removed first, and then x1. Neither x2
nor x3 were removed and the final model is ˆy = 2.18332 + 0.95758x2 + 3.32533x3.
12.50 (a) ˆy = −29.59244 + 0.27872x1 + 0.06967x2 + 1.24195x3 − 0.39554x4 + 0.22365x5.
(b) The variables x3 and x5 were entered consecutively and the final model is ˆy =
−56.94371 + 1.63447x3 + 0.24859x5.
(c) We have a summary table displayed next.
Model s2 PRESS R2 P
i |δi|
x2x5
x1x5
x1x3x5
x3x5
x3x4x5
x2x4x5
x2x3x5
x3x4
x1x2x5
x5
x3
x2x3x4x5
x1
x2x3
x1x3
x1x3x4x5
x2
x4x5
x1x2x3x5
x2x3x4
x1x2
x1x4x5
x1x3x4
x2x4
x1x2x4x5
x1x2x3x4x5
x1x4
x1x2x3
x4
x1x2x3x4
x1x2x4
176.735
174.398
174.600
194.369
192.006
196.211
186.096
249.165
184.446
269.355
257.352
197.782
274.853
264.670
226.777
188.333
328.434
289.633
195.344
269.800
297.294
192.822
240.828
352.781
207.477
214.602
287.794
249.038
613.411
266.542
317.783
2949.13
3022.18
3207.34
3563.40
3637.70
3694.97
3702.90
3803.00
3956.41
3998.77
4086.47
4131.88
4558.34
4721.55
4736.02
4873.16
4998.07
5136.91
5394.56
5563.87
5784.20
5824.58
6564.79
6902.14
7675.70
7691.30
7714.86
7752.69
8445.98
10089.94
10591.58
0.7816
0.7845
0.8058
0.7598
0.7865
0.7818
0.7931
0.6921
0.7949
0.6339
0.6502
0.8045
0.6264
0.6730
0.7198
0.8138
0.5536
0.6421
0.8069
0.7000
0.6327
0.7856
0.7322
0.5641
0.7949
0.8144
0.6444
0.7231
0.1663
0.7365
0.6466
151.681
166.223
174.819
189.881
190.564
170.847
184.285
192.172
189.107
189.373
199.520
192.000
202.533
210.853
219.630
207.542
217.814
209.232
216.934
234.565
231.374
216.583
248.123
248.621
249.604
257.732
249.221
264.324
259.968
297.640
294.044
178 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
(d) It appears that the model with x2 = LLS and x5 = Power is the best in terms of
PRESS, s2, and
P
i |δi|.
12.51 (a) ˆy = −587.21085 + 428.43313x.
(b) ˆy = 1180.00032 − 192.69121x + 35.20945x2.
(c) The summary of the two models are given as:
Model s2 R2 PRESS
μY = β0 + β1x 1,105,054 0.8378 18,811,057.08
μY = β0 + β1x + β11x2 430,712 0.9421 8,706,973.57
It appears that the model with a quadratic term is preferable.
12.52 The parameter estimate for β4 is 0.22365 with a standard error of 0.13052. Hence,
t = 1.71 with P-value = 0.6117. Fail to reject H0.
12.53 ˆσ2
b1 = 20, 588.038, ˆσ2
b11 = 62.650, and ˆσb1b11 = −1, 103.423.
12.54 (a) The following is the summary of the models.
Model s2 R2 PRESS Cp
x2x3
x2
x1x2
x1x2x3
x3
x1
x1x3
8094.15
8240.05
8392.51
8363.55
8584.27
8727.47
8632.45
0.51235
0.48702
0.49438
0.51292
0.46559
0.45667
0.47992
282194.34
282275.98
289650.65
294620.94
297242.74
304663.57
306820.37
2.0337
1.5422
3.1039
4.0000
2.8181
3.3489
3.9645
(b) The model with ln(x2) appears to have the smallest Cp with a small PRESS.
Also, the model ln(x2) and ln(x3) has the smallest PRESS. Both models appear
to better than the full model.
12.55 (a) There are many models here so the model summary is not displayed. By using
MSE criterion, the best model, contains variables x1 and x3 with s2 = 313.491.
If PRESS criterion is used, the best model contains only the constant term with
s2 = 317.51. When the Cp method is used, the best model is model with the
constant term.
(b) The normal probability plot, for the model using intercept only, is shown next.
We do not appear to have the normality.
−2 −1 0 1 2
−2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
Solutions for Exercises in Chapter 12 179
12.56 (a) [Volt = −1.64129 + 0.000556 Speed − 67.39589 Extension.
(b) P-values for the t-tests of the coefficients are all < 0.0001.
(c) The R2 = 0.9607 and the model appears to have a good fit. The residual plot
and a normal probability plot are given here.
5500 6000 6500 7000 7500 8000 8500
−400 −200 0 200 400
y^
Residual
−2 −1 0 1 2
−400 −200 0 200 400
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
12.57 (a) ˆy = 3.13682 + 0.64443x1 − 0.01042x2 + 0.50465x3 − 0.11967x4 − 2.46177x5 +
1.50441x6.
(b) The final model using the stepwise regression is
ˆy = 4.65631 + 0.51133x3 − 0.12418x4.
(c) Using Cp criterion (smaller the better), the best model is still the model stated in
(b) with s2 = 0.73173 and R2 = 0.64758. Using the s2 criterion, the model with
x1, x3 and x4 has the smallest value of 0.72507 and R2 = 0.67262. These two
models are quite competitive. However, the model with two variables has one less
variable, and thus may be more appealing.
(d) Using the model in part (b), displayed next is the Studentized residual plot. Note
that observations 2 and 14 are beyond the 2 standard deviation lines. Both of
those observations may need to be checked.
8 9 10 11 12
−2 −1 0 1 2
y^
Studentized Residual
1
2
3
4
5
6
7
8
9
10 11
12 13
14
15
16
17
18
19
12.58 The partial F-test shows a value of 0.82, with 2 and 12 degrees of freedom. Consequently,
the P-value = 0.4622, which indicates that variables x1 and x6 can be excluded
from the model.
180 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
12.59 (a) ˆy = 125.86555 + 7.75864x1 + 0.09430x2 − 0.00919x1x2.
(b) The following is the summary of the models.
Model s2 R2 PRESS Cp
x2
x1
x1x2
x1x2x3
680.00
967.91
650.14
561.28
0.80726
0.72565
0.86179
0.92045
7624.66
12310.33
12696.66
15556.11
2.8460
4.8978
3.4749
4.0000
It appears that the model with x2 alone is the best.
12.60 (a) The fitted model is ˆy = 85.75037−15.93334x1+2.42280x2+1.82754x3+3.07379x4.
(b) The summary of the models are given next.
Model s2 PRESS R2 Cp
x1x2x4
x4
x3x4
x1x2x3x4
x1x4
x2x4
x2x3x4
x1x3x4
x2
x3
x2x3
x1
x1x3
x1x2
x1x2x3
9148.76
19170.97
21745.08
10341.20
10578.94
21630.42
25160.18
12341.87
160756.81
171264.68
183701.86
95574.16
107287.63
109137.20
125126.59
447, 884.34
453, 304.54
474, 992.22
482, 210.53
488, 928.91
512, 749.78
532, 065.42
614, 553.42
1, 658, 507.38
1, 888, 447.43
1, 896, 221.30
2, 213, 985.42
2, 261, 725.49
2, 456, 103.03
2, 744, 659.14
0.9603
0.8890
0.8899
0.9626
0.9464
0.8905
0.8908
0.9464
0.0695
0.0087
0.0696
0.4468
0.4566
0.4473
0.4568
3.308
8.831
10.719
5.000
3.161
10.642
12.598
5.161
118.362
126.491
120.349
67.937
68.623
69.875
70.599
When using PRESS as well as the s2 criterion, a model with x1, x2 and x4 appears
to be the best, while when using the Cp criterion, the model with x1 and x4 is the
best. When using the model with x1, x2 and x4, we find out that the P-value for
testing β2 = 0 is 0.1980 which implies that perhaps x2 can be excluded from the
model.
(c) The model in part (b) has smaller Cp as well as competitive PRESS in comparison
to the full model.
12.61 Since H = X(X‘X)−1X‘, and
Pn
i=1
hii = tr(H), we have
Xn
i=1
hii = tr(X(X‘X)−1X‘) = tr(X‘X(X‘X)−1) = tr(Ip) = p,
where Ip is the p × p identity matrix. Here we use the property of tr(AB) = tr(BA)
in linear algebra.
Solutions for Exercises in Chapter 12 181
12.62 (a) ˆy = 9.9375 + 0.6125x1 + 1.3125x2 + 1.4625x3.
(b) The ANOVA table for all these single-degree-of-freedom components can be displayed
as:
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
x1
x2
x3
Error
1
1
1
4
3.00125
13.78125
17.11125
3.15625
0.95
4.37
5.42
0.3847
0.1049
0.0804
Total 7
Only x3 is near significant.
12.63 (a) For the completed second-order model, we have
PRESS = 9, 657, 641.55,
Xn
i=1
|yi − ˆyi,−i| = 5, 211.37.
(b) When the model does not include any term involving x4,
PRESS = 6, 954.49,
Xn
i=1
|yi − ˆyi,−i| = 277.292.
Apparently, the model without x4 is much better.
(c) For the model with x4:
PRESS = 312, 963.71,
Xn
i=1
|yi − ˆyi,−i| = 762.57.
For the model without x4:
PRESS = 3, 879.89,
Xn
i=1
|yi − ˆyi,−i| = 220.12
Once again, the model without x4 performs better in terms of PRESS and
Pn
i=1 |yi−
ˆyi,−i|.
12.64 (a) The stepwise regression results in the following fitted model:
ˆy = 102.20428 − 0.21962x1 − 0.0723 − x2 − 2.68252x3 − 0.37340x5 + 0.30491x6.
(b) Using the Cp criterion, the best model is the same as the one in (a).
12.65 (a) Yes. The orthogonality is maintained with the use of interaction terms.
182 Chapter 12 Multiple Linear Regression and Certain Nonlinear Regression Models
(b) No. There are no degrees of freedom available for computing the standard error.
12.66 The fitted model is ˆy = 26.19333+0.04772x1+0.76011x2−0.00001334x11−0.00687x22+
0.00011333x12. The t-tests for each coefficient show that x12 and x22 may be eliminated.
So, we ran a test for β12 = β22 = 0 which yields P-value = 0.2222. Therefore, both x12
and x22 may be dropped out from the model.
12.67 (a) The fitted model is ˆy = −0.26891 + 0.07755x1 + 0.02532x2 − 0.03575x3. The
f-value of the test for H0 : β1 = β2 = β3 = 0 is 35,28 with P-value < 0.0001.
Hence, we reject H0.
(b) The residual plots are shown below and they all display random residuals.
7.5 8.0 8.5 9.0 9.5 10.0
−2 −1 0 1 2
x1
Studentized Residual
0 1 2 3 4 5
−2 −1 0 1 2
x2
Studentized Residual
0 2 4 6 8
−2 −1 0 1 2
x3
Studentized Residual
(c) The following is the summary of these three models.
Model PRESS Cp
x1, x2, x3 0.091748 24.7365
x1, x2, x3, x2
1, x2
2 , x2
3 0.08446 12.3997
x1, x2, x3, x2
1 , x2
2 , x2
3, x12, x13, x23 0.088065 10
It is difficult to decide which of the three models is the best. Model I contains
all the significant variables while models II and III contain insignificant variables.
However, the Cp value and PRESS for model are not so satisfactory. Therefore,
some other models may be explored.
12.68 Denote by Z1 = 1 when Group=1, and Z1 = 0 otherwise;
Denote by Z2 = 1 when Group=2, and Z2 = 0 otherwise;
Denote by Z3 = 1 when Group=3, and Z3 = 0 otherwise;
Solutions for Exercises in Chapter 12 183
(a) The parameter estimates are:
Variable DF Parameter Estimate P-value
Intercept
BMI
z1
z2
1
1
1
1
46.34694
−1.79090
−23.84705
−17.46248
0.0525
0.0515
0.0018
0.0109
Yes, Group I has a mean change in blood pressure that was significantly lower
than the control group. It is about 23.85 points lower.
(b) The parameter estimates are:
Variable DF Parameter Estimate P-value
Intercept
BMI
z1
z3
1
1
1
1
28.88446
−1.79090
−6.38457
17.46248
0.1732
0.0515
0.2660
0.0109
Although Group I has a mean change in blood pressure that was 6.38 points lower
than that of Group II, the difference is not very significant due to a high P-value.
12.69 (b) All possible regressions should be run. R2 = 0.9908 and there is only one significant
variable.
(c) The model including x2, x3 and x5 is the best in terms of Cp, PRESS and has all
variables significant.
12.70 Using the formula of R2
adj on page 467, we have
R2
adj = 1 −
SSE/(n − k − 1)
SST/(n − 1)
= 1 −
MSE
MST
.
Since MST is fixed, maximizing R2
adj is thus equivalent to minimizing MSE.
12.71 (a) The fitted model is ˆp = 1
1+e2.7620−0.0308x .
(b) The χ2-values for testing b0 = 0 and b1 = 0 are 471.4872 and 243.4111, respectively.
Their corresponding P-values are < 0.0001. Hence, both coefficients are
significant.
(c) ED50 = −−2.7620
0.0308 = 89.675.
12.72 (a) The fitted model is ˆp = 1
1+e2.9949−0.0308x .
(b) The increase in odds of failure that results by increasing the load by 20 lb/in.2 is
e(20)(0.0308) = 1.8515.
Chapter 13
One-Factor Experiments: General
13.1 Using the formula of SSE, we have
SSE =
Xk
i=1
Xn
j=1
(yij − ¯yi.)2 =
Xk
i=1
Xn
j=1
(ǫij − ¯ǫi.)2 =
Xk
i=1
"
Xn
j=1
ǫ2
ij − n¯ǫ2
i.
#
.
Hence
E(SSE) =
Xk
i=1
"
Xn
j=1
E(ǫ2
ij) − nE(¯ǫ2
i.)
#
=
Xk
i=1
nσ2 − n
σ2
n
= k(n − 1)σ2.
Thus E
h
SSE
k(n−1)
i
= k(n−1) 2
k(n−1) = σ2.
13.2 Since SSA = n
Pk
i=1
(¯yi. − ¯y..)2 = n
Pk
i=1
¯y2
i. − kn¯y2
.., yij ∼ n(y; μ + αi, σ2), and hence
¯yi. ∼ n
y; μ + αi, √n
and ¯y.. ∼ n
μ + ¯α, √kn
, then
E(¯y2
i.) = V ar(¯yi.) + [E(¯yi.)]2 =
σ2
n
+ (μ + αi)2,
and
E[¯y2
..] =
σ2
kn
+ (μ + ¯α)2 =
σ2
kn
+ μ2,
due to the constraint on the α’s. Therefore,
E(SSA) = n
Xk
i=1
E(¯y2
i.) − knE(¯y2
..) = kσ2 + n
Xk
i=1
(μ + αi)2 − (σ2 + knμ2)
= (k − 1)σ2 + n
Xk
i=1
α2
i .
185
186 Chapter 13 One-Factor Experiments: General
13.3 The hypotheses are
H0 : μ1 = μ2 = · · · = μ6,
H1 : At least two of the means are not equal.
α = 0.05.
Critical region: f > 2.77 with v1 = 5 and v2 = 18 degrees of freedom.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatment
Error
5.34
62.64
5
18
1.07
3.48
0.31
Total 67.98 23
with P-value=0.9024.
Decision: The treatment means do not differ significantly.
13.4 The hypotheses are
H0 : μ1 = μ2 = · · · = μ5,
H1 : At least two of the means are not equal.
α = 0.05.
Critical region: f > 2.87 with v1 = 4 and v2 = 20 degrees of freedom.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Tablets
Error
78.422
59.532
4
20
19.605
2.977
6.59
Total 137.954 24
with P-value=0.0015.
Decision: Reject H0. The mean number of hours of relief differ significantly.
13.5 The hypotheses are
H0 : μ1 = μ2 = μ3,
H1 : At least two of the means are not equal.
α = 0.01.
Computation:
Solutions for Exercises in Chapter 13 187
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Shelf Height
Error
399.3
288.8
2
21
199.63
13.75
14.52
Total 688.0 23
with P-value=0.0001.
Decision: Reject H0. The amount of money spent on dog food differs with the shelf
height of the display.
13.6 The hypotheses are
H0 : μA = μB = μC,
H1 : At least two of the means are not equal.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Drugs
Error
158.867
393.000
2
27
79.433
14.556
5.46
Total 551.867 29
with P-value=0.0102.
Decision: Since α = 0.01, we fail to reject H0. However, this decision is very marginal
since the P-value is very close to the significance level.
13.7 The hypotheses are
H0 : μ1 = μ2 = μ3 = μ4,
H1 : At least two of the means are not equal.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Error
119.787
638.248
3
36
39.929
17.729
2.25
Total 758.035 39
with P-value=0.0989.
Decision: Fail to reject H0 at level α = 0.05.
188 Chapter 13 One-Factor Experiments: General
13.8 The hypotheses are
H0 : μ1 = μ2 = μ3,
H1 : At least two of the means are not equal.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Solvents
Error
3.3054
1.9553
2
29
1.6527
0.0674
24.51
Total 5.2608 31
with P-value< 0.0001.
Decision: There is significant difference in the mean sorption rate for the three solvents.
The mean sorption for the solvent Chloroalkanes is the highest. We know that it is
significantly higher than the rate of Esters.
13.9 The hypotheses are
H0 : μ1 = μ2 = μ3 = μ4,
H1 : At least two of the means are not equal.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Error
27.5506
18.6360
3
17
9.1835
1.0962
8.38
Total 46.1865 20
with P-value= 0.0012.
Decision: Reject H0. Average specific activities differ.
13.10 s50 = 3.2098, s100 = 4.5253, s200 = 5.1788, and s400 = 3.6490. Since the sample sizes
are all the same,
s2
p =
1
4
X4
i=1
s2
i = 17.7291.
Therefore, the Bartlett’s statistic is
b =
Q4
i=1
s2
i
1/4
s2
p
= 0.9335.
Solutions for Exercises in Chapter 13 189
Using Table A.10, the critical value of the Bartlett’s test with k = 4 and α = 0.05 is
0.7970. Since b > 0.7970, we fail to reject H0 and hence the variances can be assumed
equal.
13.11 Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
B vs. A, C, D
C vs. A, D
A vs. D
Error
30.6735
49.9230
5.3290
34.3800
1
1
1
16
30.6735
49.9230
5.3290
2.1488
14.28
23.23
2.48
(a) P-value=0.0016. B is significantly different from the average of A, C, and D.
(b) P-value=0.0002. C is significantly different from the average of A and D.
(c) P-value=0.1349. A can not be shown to differ significantly from D.
13.12 (a) The hypotheses are
H0 : μ29 = μ54 = μ84,
H1 : At least two of the means are not equal.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Protein Levels
Error
32, 974.87
28, 815.80
2
9
16, 487.43
3, 201.76
5.15
Total 61, 790.67 11
with P-value= 0.0323.
Decision: Reject H0. The mean nitrogen loss was significantly different for the
three protein levels.
(b) For testing the contrast L = 2μ29 − μ54 − μ84 at level α = 0.05, we have SSw =
31, 576.42 and f = 9.86, with P-value=0.0119. Hence, the mean nitrogen loss
for 29 grams of protein was different from the average of the two higher protein
levels.
13.13 (a) The hypotheses are
H0 : μ1 = μ2 = μ3 = μ4,
H1 : At least two of the means are not equal.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Error
1083.60
1177.68
3
44
361.20
26.77
13.50
Total 2261.28 47
190 Chapter 13 One-Factor Experiments: General
with P-value< 0.0001.
Decision: Reject H0. The treatment means are different.
(b) For testing two contrasts L1 = μ1 − μ2 and L2 = μ3 − μ4 at level α = 0.01, we
have the following
Contrast Sum of Squares Computed f P-value
1 vs. 2
3 vs. 4
785.47
96.00
29.35
3.59
< 0.0001
0.0648
Hence, Bath I and Bath II were significantly different for 5 launderings, and Bath
I and Bath II were not different for 10 launderings.
13.14 The means of the treatments are:
¯y1. = 5.44, ¯y2. = 7.90, ¯y3. = 4.30, ¯y4. = 2.98, and ¯y5. = 6.96.
Since q(0.05, 5, 20) = 4.24, the critical difference is (4.24)
q
2.9766
5 = 3.27. Therefore,
the Tukey’s result may be summarized as follows:
¯y5. ¯y3. ¯y1. ¯y5. ¯y2.
2.98 4.30 5.44 6.96 7.90
13.15 Since q(0.05, 4, 16) = 4.05, the critical difference is (4.05)
q
2.14875
5 = 2.655. Hence
¯y3. ¯y1. ¯y4. ¯y2.
56.52 59.66 61.12 61.96
13.16 (a) The hypotheses are
H0 : μ1 = μ2 = μ3 = μ4,
H1 : At least two of the means are not equal.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Blends
Error
119.649
44.920
3
8
39.883
5.615
7.10
Total 164.569 11
with P-value= 0.0121.
Decision: Reject H0. There is a significant difference in mean yield reduction for
the 4 preselected blends.
Solutions for Exercises in Chapter 13 191
(b) Since
p
s2/3 = 1.368 we get
p 2 3 4
rp 3.261 3.399 3.475
Rp 4.46 4.65 4.75
Therefore,
¯y3. ¯y1. ¯y2. ¯y4.
23.23 25.93 26.17 31.90
(c) Since q(0.05, 4, 8) = 4.53, the critical difference is 6.20. Hence
¯y3. ¯y1. ¯y2. ¯y4.
23.23 25.93 26.17 31.90
13.17 (a) The hypotheses are
H0 : μ1 = μ2 = · · · = μ5,
H1 : At least two of the means are not equal.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Procedures
Error
7828.30
3256.50
4
15
1957.08
217.10
9.01
Total 11084.80 19
with P-value= 0.0006.
Decision: Reject H0. There is a significant difference in the average species count
for the different procedures.
(b) Since q(0.05, 5, 15) = 4.373 and
q
217.10
4 = 7.367, the critical difference is 32.2.
Hence
¯yK ¯yS ¯ySub ¯yM ¯yD
12.50 24.25 26.50 55.50 64.25
13.18 The hypotheses are
H0 : μ1 = μ2 = · · · = μ5,
H1 : At least two of the means are not equal.
Computation:
192 Chapter 13 One-Factor Experiments: General
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Angles
Error
99.024
23.136
4
20
24.756
1.157
21.40
Total 122.160 24
with P-value< 0.0001.
Decision: Reject H0. There is a significant difference in mean pressure for the different
angles.
13.19 When we obtain the ANOVA table, we derive s2 = 0.2174. Hence
p
2s2/n =
p
(2)(0.2174)/5 = 0.2949.
The sample means for each treatment levels are
¯yC = 6.88, ¯y1. = 8.82, ¯y2. = 8.16, ¯y3. = 6.82, ¯y4. = 6.14.
Hence
d1 =
8.82 − 6.88
0.2949
= 6.579, d2 =
8.16 − 6.88
0.2949
= 4.340,
d3 =
6.82 − 6.88
0.2949
= −0.203, d4 =
6.14 − 6.88
0.2949
= −2.509.
From Table A.14, we have d0.025(4, 20) = 2.65. Therefore, concentrations 1 and 2 are
significantly different from the control.
13.20 The ANOVA table can be obtained as follows:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Cables
Error
1924.296
2626.917
8
99
240.537
26.535
9.07
Total 4551.213 107
with P-value< 0.0001.
The results from Tukey’s procedure can be obtained as follows:
¯y2. ¯y3. ¯y1. ¯y4. ¯y6. ¯y7. ¯y5. ¯y8. ¯y9.
−7.000 −6.083 −4.083 −2.667 0.833 0.917 1.917 3.333 6.250
The cables are significantly different:
9 is different from 4, 1, 2, 3
8 is different from 1, 3, 2
5, 7, 6 are different from 3, 2.
Solutions for Exercises in Chapter 13 193
13.21 Aggregate 4 has a significantly lower absorption rate than the other aggregates.
13.22 (a) The hypotheses are
H0 : μC = μL = μM = μH,
H1 : At least two of the means are not equal.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Finance Leverages
Error
80.7683
100.8700
3
20
26.9228
5.0435
5.34
Total 181.6383 23
with P-value= 0.0073.
Decision: Reject H0. The means are not all equal for the different financial
leverages.
(b)
p
2s2/n =
p
(2)(5.0435)/6 = 1.2966. The sample means for each treatment levels
are
¯yC = 4.3833, ¯yL = 5.1000, ¯yH = 8.3333, ¯yM = 8.4167.
Hence
dL =
5.1000 − 4.3833
1.2966
= 0.5528, dM =
8.4167 − 4.3833
1.2966
= 3.1108,
dL =
8.333 − 4.3833
1.2966
= 3.0464.
From Table A.14, we have d0.025(3, 20) = 2.54. Therefore, the mean rate of return
are significantly higher for median and high financial leverage than for control.
13.23 The ANOVA table can be obtained as follows:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Temperatures
Error
1268.5333
112.8333
4
25
317.1333
4.5133
70.27
Total 1381.3667 29
with P-value< 0.0001.
The results from Tukey’s procedure can be obtained as follows:
¯y0 ¯y25 ¯y100 ¯y75 ¯y50
55.167 60.167 64.167 70.500 72.833
194 Chapter 13 One-Factor Experiments: General
The batteries activated at temperature 50 and 75 have significantly longer activated
life.
13.24 The Duncan’s procedure shows the following results:
¯yE ¯yA ¯yC
0.3300 0.9422 1.0063
Hence, the sorption rate using the Esters is significantly lower than the sorption rate
using the Aromatics or the Chloroalkanes.
13.25 Based on the definition, we have the following.
SSB = k
Xb
j=1
(¯y.j − ¯y..)2 = k
Xb
j=1
T.j
k −
T..
bk
2
=
Xb
j=1
T2
.j
k − 2
T2
..
bk
+
T2
..
bk
=
Xb
j=1
T2
.j
k −
T2
..
bk
.
13.26 From the model
yij = μ + αi + βj + ǫij ,
and the constraints
Xk
i=1
αi = 0 and
Xb
j=1
βj = 0,
we obtain
¯y.j = μ + βj + ¯ǫ.j and ¯y.. = μ + ¯ǫ...
Hence
SSB = k
Xb
j=1
(¯y.j − ¯y..)2 = k
Xb
j=1
(βj + ¯ǫ.j − ¯ǫ..)2.
Since E(¯ǫ.j) = 0 and E(¯ǫ..) = 0, we obtain
E(SSB) = k
Xb
j=1
β2
j + k
Xb
j=1
E(¯ǫ2
. j) − bkE(¯ǫ2
..).
We know that E(¯ǫ2
.j) = 2
k and E(¯ǫ2
..) = 2
bk . Then
E(SSB) = k
Xb
j=1
β2
j + bσ2 − σ2 = (b − 1)σ2 + k
Xb
j=1
β2
j .
13.27 (a) The hypotheses are
H0 : α1 = α2 = α3 = α4 = 0, fertilizer effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Critical region: f > 4.76.
Computation:
Solutions for Exercises in Chapter 13 195
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Fertilizers
Blocks
Error
218.1933
197.6317
71.4017
3
2
6
72.7311
98.8158
11.9003
6.11
Total 487.2267 11
P-value= 0.0296. Decision: Reject H0. The means are not all equal.
(b) The results of testing the contrasts are shown as:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
(f1, f3) vs (f2. f4)
f1 vs f3
Error
206.6700
11.4817
71.4017
1
1
6
206.6700
11.4817
11.9003
17.37
0.96
The corresponding P-values for the above contrast tests are 0.0059 and 0.3639,
respectively. Hence, for the first contrast, the test is significant and the for the
second contrast, the test is insignificant.
13.28 The hypotheses are
H0 : α1 = α2 = α3 = 0, no differences in the varieties
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Critical region: f > 5.14.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Blocks
Error
24.500
171.333
42.167
2
3
6
12.250
57.111
7.028
1.74
Total 238.000 11
P-value=0.2535. Decision: Do not reject H0; could not show that the varieties of
potatoes differ in field.
13.29 The hypotheses are
H0 : α1 = α2 = α3 = 0, brand effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Critical region: f > 3.84.
Computation:
196 Chapter 13 One-Factor Experiments: General
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Blocks
Error
27.797
16.536
18.556
2
4
8
13.899
4.134
2.320
5.99
Total 62.889 14
P-value=0.0257. Decision: Reject H0; mean percent of foreign additives is not the
same for all three brand of jam. The means are:
Jam A: 2.36, Jam B: 3.48, Jam C: 5.64.
Based on the means, Jam A appears to have the smallest amount of foreign additives.
13.30 The hypotheses are
H0 : α1 = α2 = α3 = α4 = 0, courses are equal difficulty
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Subjects
Students
Error
42.150
1618.700
1112.100
3
4
12
14.050
404.675
92.675
0.15
Total 2772.950 19
P-value=0.9267. Decision: Fail to reject H0; there is no significant evidence to conclude
that courses are of different difficulty.
13.31 The hypotheses are
H0 : α1 = α2 = · · · = α6 = 0, station effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Stations
Dates
Error
230.127
3.259
44.018
5
5
25
46.025
0.652
1.761
26.14
Total 277.405 35
Solutions for Exercises in Chapter 13 197
P-value< 0.0001. Decision: Reject H0; the mean concentration is different at the
different stations.
13.32 The hypotheses are
H0 : α1 = α2 = α3 = 0, station effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Stations
Months
Error
10.115
537.030
744.416
2
11
22
5.057
48.821
33.837
0.15
Total 1291.561 35
P-value= 0.8620. Decision: Do not reject H0; the treatment means do not differ
significantly.
13.33 The hypotheses are
H0 : α1 = α2 = α3 = 0, diet effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Diets
Subjects
Error
4297.000
6033.333
1811.667
2
5
10
2148.500
1206.667
181.167
11.86
Total 12142.000 17
P-value= 0.0023. Decision: Reject H0; differences among the diets are significant.
13.34 The hypotheses are
H0 : α1 = α2 = α3 = 0, analyst effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Computation:
198 Chapter 13 One-Factor Experiments: General
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Analysts
Individuals
Error
0.001400
0.021225
0.001400
2
3
6
0.000700
0.007075
0.000233
3.00
Total 0.024025 11
P-value= 0.1250. Decision: Do not reject H0; cannot show that the analysts differ
significantly.
13.35 The hypotheses are
H0 : α1 = α2 = α3 = α4 = α5 = 0, treatment effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Locations
Error
79630.133
634334.667
689106.667
4
5
20
19907.533
126866.933
34455.333
0.58
Total 1403071.467 29
P-value= 0.6821. Decision: Do not reject H0; the treatment means do not differ
significantly.
13.36 The hypotheses are
H0 : α1 = α2 = α3 = 0, treatment effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Subjects
Error
203.2792
188.2271
212.8042
2
9
18
101.6396
20.9141
11.8225
8.60
Total 604.3104 29
P-value= 0.0024. Decision: Reject H0; the mean weight losses are different for different
treatments and the therapists had the greatest effect on the weight loss.
Solutions for Exercises in Chapter 13 199
13.37 The total sums of squares can be written as
X
i
X
j
X
k
(yijk − ¯y...)2 =
X
i
X
j
X
k
[(¯yi.. − ¯y...) + (¯y.j. − ¯y...) + (¯y..k − ¯y...)
+ (yijk − ¯yi.. − ¯y.j. − ¯y..k + 2¯y...)]2
=r
X
i
(¯yi.. − ¯y...)2 + r
X
j
(¯y.j. − ¯y...)2 + r
X
k
(¯y..k − ¯y...)2
+
X
i
X
j
X
k
(yijk − ¯yi.. − ¯y.j. − ¯y..k + 2¯y...)2
+ 6 cross-product terms,
and all cross-product terms are equal to zeroes.
13.38 For the model yijk = μ + αi + βj + τk + ǫijk, we have
¯y..k = μ + τk + ¯ǫ..k, and ¯y... = μ + ¯ǫ....
Hence SSTr = r
P
k
(¯y..k − ¯y...)2 = r
P
k
(τk + ¯ǫ..k − ¯ǫ...)2, and
E(SSTr) = r
X
k
τ 2
k + r
X
k
E(¯ǫ2
..k) − r2
X
k
E(¯ǫ2
...)
= r
X
k
τ 2
k + r
X
k
σ2
r − r2 σ2
r2 = r
X
k
τ 2
k + rσ2 − σ2
= (r − 1)σ2 + r
X
k
τ 2
k .
13.39 The hypotheses are
H0 : τ1 = τ2 = τ3 = τ4 = 0, professor effects are zero
H1 : At least one of the τi’s is not equal to zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Time Periods
Courses
Professors
Error
474.50
252.50
723.50
287.50
3
3
3
6
158.17
84.17
241.17
47.92
5.03
Total 1738.00 15
P-value= 0.0446. Decision: Reject H0; grades are affected by different professors.
200 Chapter 13 One-Factor Experiments: General
13.40 The hypotheses are
H0 : τA = τB = τC = τD = τE = 0, color additive effects are zero
H1 : At least one of the τi’s is not equal to zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Workers
Days
Additives
Error
12.4344
14.7944
3.9864
9.2712
4
4
4
12
3.1086
3.6986
0.9966
0.7726
1.29
Total 40.4864 24
P-value= 0.3280. Decision: Do not reject H0; color additives could not be shown to
have an effect on setting time.
13.41 The hypotheses are
H0 : α1 = α2 = α3 = 0, dye effects are zero
H1 : At least one of the αi’s is not equal to zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Amounts
Plants
Error
1238.8825
53.7004
101.2433
2
1
20
619.4413
53.7004
5.0622
122.37
Total 1393.8263 23
P-value< 0.0001. Decision: Reject H0; color densities of fabric differ significantly for
three levels of dyes.
13.42 (a) After a transformation g(y) = √y, we come up with an ANOVA table as follows.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Materials
Error
7.5123
6.2616
2
27
3.7561
0.2319
16.20
Total 13.7739 29
Solutions for Exercises in Chapter 13 201
(b) The P-value< 0.0001. Hence, there is significant difference in flaws among three
materials.
(c) A residual plot is given below and it does show some heterogeneity of the variance
among three treatment levels.
1.0 1.5 2.0 2.5 3.0
−0.5 0.0 0.5 1.0
material
residual
(d) The purpose of the transformation is to stabilize the variances.
(e) One could be the distribution assumption itself. Once the data is transformed, it
is not necessary that the data would follow a normal distribution.
(f) Here the normal probability plot on residuals is shown.
−2 −1 0 1 2
−0.5 0.0 0.5 1.0
Theoretical Quantiles
Sample Quantiles
It appears to be close to a straight
line. So, it is likely that the transformed data are normally distributed.
13.43 (a) The hypotheses are
H0 : σ2
= 0,
H1 : σ2
6= 0
α = 0.05.
Computation:
202 Chapter 13 One-Factor Experiments: General
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Operators
Error
371.8719
99.7925
3
12
123.9573
8.3160
14.91
Total 471.6644 15
P-value= 0.0002. Decision: Reject H0; operators are different.
(b) ˆσ2 = 8.316 and ˆσ2
= 123.9573−8.3160
4 = 28.910.
13.44 The model is yij = μ + Ai + Bj + ǫij . Hence
¯y.j = μ + ¯ A. + Bj + ¯ǫ.j , and ¯y.. = μ + ¯ A. + ¯B. + ¯ǫ...
Therefore,
SSB = k
Xb
j=1
(¯y.j − ¯y..)2 = k
Xb
j=1
[(Bj − ¯B.) + (¯ǫ.j − ¯ǫ..)]2,
and
E(SSB) = k
Xb
j=1
E(B2
j ) − kbE(¯B2
. ) + k
Xb
j=1
E(¯ǫ2
.j) − kbE(¯ǫ2
..)
= kbσ2
− kσ2
+ bσ2 − σ2 = (b − 1)σ2 + k(b − 1)σ2
.
13.45 (a) The hypotheses are
H0 : σ2
= 0,
H1 : σ2
6= 0
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Treatments
Blocks
Error
23.238
45.283
27.937
3
4
12
7.746
11.321
2.328
3.33
Total 96.458 19
P-value= 0.0565. Decision: Not able to show a significant difference in the random
treatments at 0.05 level, although the P-value shows marginal significance.
(b) σ2
= 7.746−2.328
5 = 1.084, and σ2
= 11.321−2.328
4 = 2.248.
13.46 From the model
yijk = μ + Ai + Bj + Tk + ǫij ,
Solutions for Exercises in Chapter 13 203
we have
¯y..k = μ + ¯ A. + ¯B. + Tk + ¯ǫ..k, and ¯y... = μ + ¯ A. + ¯B. + ¯ T. + ¯ǫ....
Hence,
SSTr = r
X
k
(¯y..k − ¯y...)2 = r
X
k
[(Tj − ¯ T.) + (¯ǫ..k − ¯ǫ...)]2,
and
E(SSTr) = r
X
k
E(T2
k ) − r2E( ¯ T2
. ) + r
X
k
E(¯ǫ2
. .k) − r2E(¯ǫ2
...)
= r2σ2
− rσ2
+ rσ2 − σ2 = (r − 1)(σ2 + rσ2
).
13.47 (a) The matrix is
A =
bk b b · · · b k k · · · k
b b 0 · · · 0 1 1 · · · 1
b 0 b · · · 0 1 1 · · · 1
...
...
...
. . .
...
...
...
. . .
...
b 0 0 · · · b 1 1 · · · 1
k 1 1 · · · k 0 0 · · · 0
k 1 1 · · · 0 k 0 · · · 0
...
...
...
. . .
...
...
...
. . .
...
k 1 1 · · · 0 0 0 · · · k
,
where b = number of blocks and k = number of treatments. The vectors are
b
′
= (μ, α1, α2, · · · , αk, β1, β2, · · · , βb)′, and
g
′
= (T.., T1., T2., · · · , Tk., T.1, T.2, · · · , T.b)
′
.
(b) Solving the system Ab = g with the constraints
Pk
i=1
αi = 0 and
Pb
j=1
βj = 0, we
have
ˆμ = ¯y..,
ˆαi = ¯yi. − ¯y.., for i = 1, 2, . . . , k,
ˆ βj = ¯y.j − ¯y.., for j = 1, 2, . . . , b.
Therefore,
R(α1, α2, . . . , αk, β1, β2, . . . , βb) = b
′
g −
T2
..
bk
=
Xk
i=1
T2
i.
b
+
Xb
j=1
T2
.j
k − 2
T2
..
bk
.
204 Chapter 13 One-Factor Experiments: General
To find R(β1, β2, . . . , βb | α1, α2, . . . , αk) we first find R(α1, α2, . . . , αk). Setting
βj = 0 in the model, we obtain the estimates (after applying the constraint
Pk
i=1
αi = 0)
ˆμ = ¯y.., and ˆαi = ¯yi. − ¯y.., for i = 1, 2, . . . , k.
The g vector is the same as in part (a) with the exception that T.1, T.2, . . . , T.b do
not appear. Thus one obtains
R(α1, α2, . . . , αk) =
Xk
i=1
T2
i.
b −
T2
..
bk
and thus
R(β1, β2, . . . , βb | α1, α2, . . . , αk) = R(α1, α2, . . . , αk, β1, β2, . . . , βb)
− R(α1, α2, . . . , αk) =
Xb
j=1
T2
.j
k −
T2
..
bk
= SSB.
13.48 Since
1 − β = P
F(3, 12) >
3.49
1 + (4)(1.5)
= P[F(12, 3) < 2.006] < 0.95.
Hence we do not have large enough samples. We then find, by trial and error, that
n = 16 is sufficient since
1 − β = P
F(3, 60) >
2.76
1 + (16)(1.5)
= P[F(60, 3) < 9.07] > 0.95.
13.49 We know φ2 = b
P4
i=1
2
i
4 2 = b
2 , when
P4
i=1
2
i
2 = 2.0.
If b = 10, φ = 2.24; v1 = 3 and v2 = 27 degrees of freedom.
If b = 9, φ = 2.12; v1 = 3 and v2 = 24 degrees of freedom.
If b = 8, φ = 2.00; v1 = 3 and v2 = 21 degrees of freedom.
From Table A.16 we see that b = 9 gives the desired result.
13.50 For the randomized complete block design we have
E(S2
1 ) = E
SSA
k − 1
= σ2 + b
Xk
i=1
α2
i
k − 1
.
Therefore,
λ =
v1[E(S2
1 )]
2σ2 −
v1
2
=
(k − 1)
σ2 + b
Pk
i=1
α2
i /(k − 1)
2σ2 −
k − 1
2
= b
Xk
i=1
α2
i
2σ2 ,
Solutions for Exercises in Chapter 13 205
and then
φ2 =
E(S2
1 ) − σ2
σ2 ·
v1
v1 + 1
=
[σ2 + b
Pk
i=1
α2
i /(k − 1)] − σ2
σ2 ·
k − 1
k
= b
Xk
i=1
α2
i
kσ2 .
13.51 (a) The model is yij = μ + αi + ǫij , where αi ∼ n(0, σ2
).
(b) Since s2 = 0.02056 and s2
1 = 0.01791, we have ˆσ2 = 0.02056 and s21
−s2
10 =
0.01791−0.02056
10 = −0.00027, which implies ˆσ2
= 0.
13.52 (a) The P-value of the test result is 0.7830. Hence, the variance component of pour
is not significantly different from 0.
(b) We have the ANOVA table as follows:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Pours
Error
0.08906
1.02788
4
20
0.02227
0.05139
0.43
Total 1.11694 24
Since s21
−s2
5 = 0.02227−0.05139
5 < 0, we have ˆσ2
= 0.
13.53 (a) yij = μ + αi + ǫij , where αi ∼ n(x; 0, σ2
).
(b) Running an ANOVA analysis, we obtain the P-value as 0.0121. Hence, the loom
variance component is significantly different from 0 at level 0.05.
(c) The suspicion is supported by the data.
13.54 The hypotheses are
H0 : μ1 = μ2 = μ3 = μ4,
H1 : At least two of the μi’s are not equal.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Garlon levels
Error
3.7289
9.5213
3
12
1.2430
0.7934
1.57
Total 13.2502 15
P-value= 0.2487. Decision: Do not reject H0; there is insufficient evidence to claim
that the concentration levels of Garlon would impact the heights of shoots.
206 Chapter 13 One-Factor Experiments: General
13.55 Bartlett’s statistic is b = 0.8254. Conclusion: do not reject homogeneous variance
assumption.
13.56 The hypotheses are
H0 : τA = τB = τC = τD = τE = 0, ration effects are zero
H1 : At least one of the τi’s is not zero.
α = 0.01.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Lactation Periods
Cows
Rations
Error
245.8224
353.1224
89.2624
21.3392
4
4
4
12
61.4556
88.2806
22.3156
1.7783
12.55
Total 709.5464 24
P-value= 0.0003. Decision: Reject H0; different rations have an effect on the daily
milk production.
13.57 It can be shown that ¯yC = 76.16, ¯y1 = 81.20, ¯y2 = 81.54 and ¯y3 = 80.98. Since this is
a one-sided test, we find d0.01(3, 16) = 3.05 and
r
2s2
n
=
r
(2)(3.52575)
5
= 1.18756.
Hence
d1 =
81.20 − 76.16
1.18756
= 4.244, d2 =
81.54 − 76.16
1.18756
= 4.532, d3 =
80.98 − 76.16
1.18756
= 4.059,
which are all larger than the critical value. Hence, significantly higher yields are
obtained with the catalysts than with no catalyst.
13.58 (a) The hypotheses for the Bartlett’s test are
H0 : σ2
A = σ2
B = σ2
C = σ2
D ,
H1 : The variances are not all equal.
α = 0.05.
Critical region: We have n1 = n2 = n3 = n4 = 5, N = 20, and k = 4. Therefore,
we reject H0 when b < b4(0.05, 5) = 0.5850.
Computation: sA = 1.40819, sB = 2.16056, sC = 1.16276, sD = 0.76942 and
hence sp = 1.46586. From these, we can obtain that b = 0.7678.
Decision: Do not reject H0; there is no sufficient evidence to conclude that the
variances are not equal.
Solutions for Exercises in Chapter 13 207
(b) The hypotheses are
H0 : μA = μB = μC = μD,
H1 : At least two of the μi’s are not equal.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Laboratories
Error
85.9255
34.3800
3
16
28.6418
2.1488
13.33
Total 120.3055 19
P-value= 0.0001. Decision: Reject H0; the laboratory means are significantly
different.
(c) The normal probability plot is given as follows:
−2 −1 0 1 2
−2 −1 0 1 2
Theoretical Quantiles
Sample Quantiles
13.59 The hypotheses for the Bartlett’s test are
H0 : σ2
1 = σ2
2 = σ2
3 = σ2
4,
H1 : The variances are not all equal.
α = 0.01.
Critical region: We have n1 = n2 = n3 = 4, n4 = 9 N = 21, and k = 4. Therefore, we
reject H0 when
b < b4(0.01, 4, 4, 4, 9)
=
(4)(0.3475) + (4)(0.3475) + (4)(0.3475) + (9)(0.6892)
21
= 0.4939.
Computation: s2
1 = 0.41709, s2
2 = 0.93857, s2
3 = 0.25673, s2
4 = 1.72451 and hence
s2
p = 1.0962. Therefore,
b =
[(0.41709)3(0.93857)3(0.25763)3(1.72451)8]1/17
1.0962
= 0.79.
208 Chapter 13 One-Factor Experiments: General
Decision: Do not reject H0; the variances are not significantly different.
13.60 The hypotheses for the Cochran’s test are
H0 : σ2
A = σ2
B = σ2
C ,
H1 : The variances are not all equal.
α = 0.01.
Critical region: g > 0.6912.
Computation: s2
A = 29.5667, s2
B = 10.8889, s2
C = 3.2111, and hence
P
s2
i = 43.6667.
Now, g = 29.5667
43.6667 = 0.6771.
Decision: Do not reject H0; the variances are not significantly different.
13.61 The hypotheses for the Bartlett’s test are
H0 : σ2
1 = σ2
2 = σ2
3,
H1 : The variances are not all equal.
α = 0.05.
Critical region: reject H0 when
b < b4(0.05, 9, 8, 15) =
(9)(0.7686) + (8)(0.7387) + (15)(0.8632)
32
= 0.8055.
Computation: b = [(0.02832)8(0.16077)7(0.04310)14]1/29
0.067426 = 0.7822.
Decision: Reject H0; the variances are significantly different.
13.62 (a) The hypotheses are
H0 : α1 = α2 = α3 = α4 = 0,
H1 : At least one of the αi’s is not zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Diets
Blocks
Error
822.1360
17.1038
106.3597
3
5
15
274.0453
3.4208
7.0906
38.65
Total 945.5995 23
with P-value< 0.0001.
Decision: Reject H0; diets do have a significant effect on mean percent dry matter.
Solutions for Exercises in Chapter 13 209
(b) We know that ¯yC = 35.8483, ¯yF = 36.4217, ¯yT = 45.1833, ¯yA = 49.6250, and
r
2s2
n
=
r
(2)(7.0906)
6
= 1.5374.
Hence,
dAmmonia =
49.6250 − 35.8483
1.5374
= 8.961,
dUrea Feeding =
36.4217 − 35.8483
1.5374
= 0.3730,
dUrea Treated =
45.1833 − 35.8483
1.5374
= 6.0719.
Using the critical value d0.05(3, 15) = 2.61, we obtain that only “Urea Feeding” is
not significantly different from the control, at level 0.05.
(c) The normal probability plot for the residuals are given below.
−2 −1 0 1 2
−4 −2 0 2 4
Theoretical Quantiles
Sample Quantiles
13.63 The hypotheses are
H0 : α1 = α2 = α3 = 0,
H1 : At least one of the αi’s is not zero.
α = 0.05.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Diet
Error
0.32356
0.20808
2
12
0.16178
0.01734
9.33
Total 0.53164 14
210 Chapter 13 One-Factor Experiments: General
with P-value= 0.0036.
Decision: Reject H0; zinc is significantly different among the diets.
13.64 (a) The gasoline manufacturers would want to apply their results to more than one
model of car.
(b) Yes, there is a significant difference in the miles per gallon for the three brands
of gasoline.
(c) I would choose brand C for the best miles per gallon.
13.65 (a) The process would include more than one stamping machine and the results might
differ with different machines.
(b) The mean plot is shown below.
1
1
1
3.2 3.4 3.6 3.8 4.0 4.2 4.4
Material
Number of Gaskets
2
2
2
cork plastic rubber
(c) Material 1 appears to be the best.
(d) Yes, there is interaction. Materials 1 and 3 have better results with machine 1
but material 2 has better results with machine 2.
13.66 (a) The hypotheses are
H0 : α1 = α2 = α3 = 0,
H1 : At least one of the αi’s is not zero.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Paint Types
Error
227875.11
336361.83
2
15
113937.57
22424.12
5.08
Total 564236.94 17
with P-value= 0.0207.
Decision: Reject H0 at level 0.05; the average wearing quality differs significantly
for three paints.
(b) Using Tukey’s test, it turns out the following.
Solutions for Exercises in Chapter 13 211
¯y1. ¯y3. ¯y2.
197.83 419.50 450.50
Types 2 and 3 are not significantly different, while Type 1 is significantly different
from Type 2.
(c) We plot the residual plot and the normal probability plot for the residuals as
follows.
1.0 1.5 2.0 2.5 3.0
−200 −100 0 100 200
Type
Residual
−2 −1 0 1 2
−200 −100 0 100 200
Theoretical Quantiles
Sample Quantiles
It appears that the heterogeneity in variances may be violated, as is the normality
assumption.
(d) We do a log transformation of the data, i.e., y′ = log(y). The ANOVA result has
changed as follows.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Paint Types
Error
2.6308
3.2516
2
15
1.3154
0.2168
6.07
Total 5.8824 17
with P-value= 0.0117.
Decision: Reject H0 at level 0.05; the average wearing quality differ significantly
for three paints. The residual and normal probability plots are shown here:
1.0 1.5 2.0 2.5 3.0
−0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 0.6
Type
Residual −
2
−
1
0
1
2
−0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 0.6
Theoretical Quantiles
Sample Quantiles
While the homogeneity of the variances seem to be a little better, the normality
assumption may still be invalid.
212 Chapter 13 One-Factor Experiments: General
13.67 (a) The hypotheses are
H0 : α1 = α2 = α3 = α4 = 0,
H1 : At least one of the αi’s is not zero.
Computation:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Locations
Error
0.01594
0.00616
3
16
0.00531
0.00039
13.80
Total 0.02210 19
with P-value= 0.0001.
Decision: Reject H0; the mean ozone levels differ significantly across the locations.
(b) Using Tukey’s test, the results are as follows.
¯y4. ¯y1. ¯y3. ¯y2.
0.078 0.092 0.096 0.152
Location 2 appears to have much higher ozone measurements than other locations.
Chapter 14
Factorial Experiments (Two or More
Factors)
14.1 The hypotheses of the three parts are,
(a) for the main effects temperature,
H
′
0 : α1 = α2 = α3 = 0,
H
′
1 : At least one of the αi’s is not zero;
(b) for the main effects ovens,
H
′′
0 : β1 = β2 = β3 = β4 = 0,
H
′′
1 : At least one of the βi’s is not zero;
(c) and for the interactions,
H
′′′
0 : (αβ)11 = (αβ)12 = · · · = (αβ)34 = 0,
H
′′′
1 : At least one of the (αβ)ij ’s is not zero.
α = 0.05.
Critical regions: (a) f1 > 3.00; (b) f2 > 3.89; and (c) f3 > 3.49.
Computations: From the computer printout we have
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Temperatures
Ovens
Interaction
Error
5194.08
4963.12
3126.26
3833.50
2
3
6
12
2597.0400
1654.3733
521.0433
319.4583
8.13
5.18
1.63
Total 17, 116.96 23
213
214 Chapter 14 Factorial Experiments (Two or More Factors)
Decision: (a) Reject H′
0; (b) Reject H′′
0 ; (c) Do not reject H′′′
0 .
14.2 The hypotheses of the three parts are,
(a) for the main effects brands,
H
′
0 : α1 = α2 = α3 = 0,
H
′
1 : At least one of the αi’s is not zero;
(b) for the main effects times,
H
′′
0 : β1 = β2 = β3 = 0,
H
′′
1 : At least one of the βi’s is not zero;
(c) and for the interactions,
H
′′′
0 : (αβ)11 = (αβ)12 = · · · = (αβ)33 = 0,
H
′′′
1 : At least one of the (αβ)ij ’s is not zero.
α = 0.05.
Critical regions: (a) f1 > 3.35; (b) f2 > 3.35; and (c) f3 > 2.73.
Computations: From the computer printout we have
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Brands
Times
Interaction
Error
32.7517
227.2117
17.3217
254.7025
2
2
4
27
16.3758
113.6058
4.3304
9.4334
1.74
12.04
0.46
Total 531.9875 35
Decision: (a) Do not reject H′
0; (b) Reject H′′
0 ; (c) Do not reject H′′′
0 .
14.3 The hypotheses of the three parts are,
(a) for the main effects environments,
H
′
0 : α1 = α2 = 0, (no differences in the environment)
H
′
1 : At least one of the αi’s is not zero;
(b) for the main effects strains,
H
′′
0 : β1 = β2 = β3 = 0, (no differences in the strains)
H
′′
1 : At least one of the βi’s is not zero;
Solutions for Exercises in Chapter 14 215
(c) and for the interactions,
H
′′′
0 : (αβ)11 = (αβ)12 = · · · = (αβ)23 = 0, (environments and strains do not interact)
H
′′′
1 : At least one of the (αβ)ij ’s is not zero.
α = 0.01.
Critical regions: (a) f1 > 7.29; (b) f2 > 5.16; and (c) f3 > 5.16.
Computations: From the computer printout we have
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Environments
Strains
Interaction
Error
14, 875.521
18, 154.167
1, 235.167
42, 192.625
1
2
2
42
14, 875.521
9, 077.083
617.583
1004.586
14.81
9.04
0.61
Total 76, 457.479 47
Decision: (a) Reject H′
0; (b) Reject H′′
0 ; (c) Do not reject H′′′
0 . Interaction is not
significant, while both main effects, environment and strain, are all significant.
14.4 (a) The hypotheses of the three parts are,
H
′
0 : α1 = α2 = α3 = 0
H
′
1 : At least one of the αi’s is not zero;
H
′′
0 : β1 = β2 = β3 = 0,
H
′′
1 : At least one of the βi’s is not zero;
H
′′′
0 : (αβ)11 = (αβ)12 = · · · = (αβ)33 = 0,
H
′′′
1 : At least one of the (αβ)ij ’s is not zero.
α = 0.01.
Critical regions: for H′
0, f1 > 3.21; for H′′
0 , f2 > 3.21; and for H′′′
0 , f3 > 2.59.
Computations: From the computer printout we have
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Coating
Humidity
Interaction
Error
1, 535, 021.37
1, 020, 639.15
1, 089, 989.63
5, 028, 396.67
2
2
4
45
767, 510.69
510, 319.57
272, 497.41
111, 742.15
6.87
4.57
2.44
Total 76, 457.479 47
Decision: Reject H′
0; Reject H′′
0 ; Do not reject H′′′
0 . Coating and humidity do not
interact, while both main effects are all significant.
216 Chapter 14 Factorial Experiments (Two or More Factors)
(b) The three means for the humidity are ¯yL = 733.78, ¯yM = 406.39 and ¯yH = 638.39.
Using Duncan’s test, the means can be grouped as
¯yM ¯yL ¯yH
406.39 638.39 733.78
Therefore, corrosion damage is different for medium humidity than for low or high
humidity.
14.5 The hypotheses of the three parts are,
(a) for the main effects subjects,
H
′
0 : α1 = α2 = α3 = 0,
H
′
1 : At least one of the αi’s is not zero;
(b) for the main effects muscles,
H
′′
0 : β1 = β2 = β3 = β4 = β5 = 0,
H
′′
1 : At least one of the βi’s is not zero;
(c) and for the interactions,
H
′′′
0 : (αβ)11 = (αβ)12 = · · · = (αβ)35 = 0,
H
′′′
1 : At least one of the (αβ)ij ’s is not zero.
α = 0.01.
Critical regions: (a) f1 > 5.39; (b) f2 > 4.02; and (c) f3 > 3.17.
Computations: From the computer printout we have
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f
Subjects
Muscles
Interaction
Error
4, 814.74
7, 543.87
11, 362.20
2, 099.17
2
4
8
30
2, 407.37
1, 885.97
1, 420.28
69.97
34.40
26.95
20.30
Total 25, 819.98 44
Decision: (a) Reject H′
0; (b) Reject H′′
0 ; (c) Reject H′′′
0 .
14.6 The ANOVA table is shown as
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Additive
Temperature
Interaction
Error
1.7578
0.8059
1.7934
1.8925
1
3
3
24
1.7578
0.2686
0.5978
0.0789
22.29
3.41
7.58
< 0.0001
0.0338
0.0010
Total 6.2497 32
Solutions for Exercises in Chapter 14 217
Decision: All main effects and interaction are significant.
An interaction plot is given here.
1
1
3.0 3.2 3.4 3.6 3.8 4.0
Additive
Adhesiveness
2
2
3
3
4
4
0 1
Temperature
1234
50
60
70
80
14.7 The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Temperature
Catalyst
Interaction
Error
430.475
2, 466.650
326.150
264.500
3
4
12
20
143.492
616.663
27.179
13.225
10.85
46.63
2.06
0.0003
< 0.0001
0.0745
Total 3, 487.775 39
Decision: All main effects are significant and the interaction is significant at level
0.0745. Hence, if 0.05 significance level is used, interaction is not significant.
An interaction plot is given here.
1
1
1
1
1
40 45 50 55 60 65 70
Amount of Catalyst
Extraction Rate
2
2
2
2
2
3
3
3
3
3
4
4 4
4
4
0.5 0.6 0.7 0.8 0.9
Temperature
1234
50
60
70
80
Duncan’s tests, at level 0.05, for both main effects result in the following.
(a) For Temperature:
¯y50 ¯y80 ¯y70 ¯y60
52.200 59.000 59.800 60.300
218 Chapter 14 Factorial Experiments (Two or More Factors)
(b) For Amount of Catalyst:
¯y0.5 ¯y0.6 ¯y0.9 ¯y0.7 ¯y0.8
44.125 56.000 58.125 64.625 66.250
14.8 (a) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Nickel
pH
Nickel*pH
Error
31, 250.00
6, 606.33
670.33
8, 423.33
1
2
2
12
31, 250.00
3, 303.17
335.17
701.94
44.52
4.71
0.48
< 0.0001
0.0310
0.6316
Total 3, 487.775 39
Decision: Nickel contents and levels of pH do not interact to each other, while
both main effects of nickel contents and levels of pH are all significant, at level
higher than 0.0310.
(b) In comparing the means of the six treatment combinations, a nickel content level
of 18 and a pH level of 5 resulted in the largest values of thickness.
(c) The interaction plot is given here and it shows no apparent interactions.
1
1
1
80 100 120 140 160 180 200
pH
Thickness
2
2 2
5 5.5 6
Nickel
12
10
18
14.9 (a) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Tool
Speed
Tool*Speed
Error
675.00
12.00
192.00
72.67
1
1
1
8
675.00
12.00
192.00
9.08
74.31
1.32
21.14
< 0.0001
0.2836
0.0018
Total 951.67 11
Decision: The interaction effects are significant. Although the main effects of
speed showed insignificance, we might not make such a conclusion since its effects
might be masked by significant interaction.
Solutions for Exercises in Chapter 14 219
(b) In the graph shown, we claim that the cutting speed that results in the longest
life of the machine tool depends on the tool geometry, although the variability of
the life is greater with tool geometry at level 1.
1
1
5 10 15 20 25 30 35
Tool Geometry
Life
2
2
1 2
Speed
12
High
Low
(c) Since interaction effects are significant, we do the analysis of variance for separate
tool geometry.
(i) For tool geometry 1, an f-test for the cutting speed resulted in a P-value =
0.0405 with the mean life (standard deviation) of the machine tool at 33.33
(4.04) for high speed and 23.33 (4.16) for low speed. Hence, a high cutting
speed has longer life for tool geometry 1.
(ii) For tool geometry 2, an f-test for the cutting speed resulted in a P-value =
0.0031 with the mean life (standard deviation) of the machine tool at 10.33
(0.58) for high speed and 16.33 (1.53) for low speed. Hence, a low cutting
speed has longer life for tool geometry 2.
For the above detailed analysis, we note that the standard deviations for the mean
life are much higher at tool geometry 1.
(d) See part (b).
14.10 (a) yijk = μ + αi + βj + (αβ)ij + ǫijk, i = 1, 2, . . . , a; j = 1, 2, . . . , b; k = 1, 2, . . . , n.
(b) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Dose
Position
Error
117.9267
15.0633
13.0433
1
2
2
117.9267
7.5317
6.5217
18.08
1.15
0.0511
0.4641
Total 146.0333 5
(c) (n − 1) − (a − 1) − (b − 1) = 5 − 1 − 2 = 2.
(d) At level 0.05, Tukey’s result for the furnace position is shown here:
¯y2 ¯y1 ¯y3
19.850 21.350 23.700
220 Chapter 14 Factorial Experiments (Two or More Factors)
Although Tukey’s multiple comparisons resulted in insignificant differences among
the furnace position levels, based on a P-value of 0.0511 for the Dose and on the
plot shown we can see that Dose=2 results in higher resistivity.
1
1
1
16 18 20 22 24 26
Furnace Position
Resistivity
2
2
2
1 2 3
Dose
12
12
14.11 (a) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Method
Lab
Method*Lab
Error
0.000104
0.008058
0.000198
0.000222
1
6
6
14
0.000104
0.001343
0.000033
0.000016
6.57
84.70
2.08
0.0226
< 0.0001
0.1215
Total 0.00858243 27
(b) Since the P-value = 0.1215 for the interaction, the interaction is not significant.
Hence, the results on the main effects can be considered meaningful to the scientist.
(c) Both main effects, method of analysis and laboratory, are all significant.
(d) The interaction plot is show here.
1
1
1
1
1
1
1
0.10 0.11 0.12 0.13 0.14 0.15 0.16
Lab
Sulfur Percent
2
2
2
2
2
2
2
1 2 3 4 5 6 7
Method
12
12
(e) When the tests are done separately, i.e., we only use the data for Lab 1, or Lab
2 alone, the P-values for testing the differences of the methods at Lab 1 and 7
Solutions for Exercises in Chapter 14 221
are 0.8600 and 0.1557, respectively. In this case, usually the degrees of freedom
of errors are small. If we compare the mean differences of the method within the
overall ANOVA model, we obtain the P-values for testing the differences of the
methods at Lab 1 and 7 as 0.9010 and 0.0093, respectively. Hence, methods are
no difference in Lab 1 and are significantly different in Lab 7. Similar results may
be found in the interaction plot in (d).
14.12 (a) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Time
Copper
Time*Copper
Error
0.025622
0.008956
0.012756
0.005933
2
2
4
18
0.012811
0.004478
0.003189
0.000330
38.87
13.58
9.67
< 0.0001
0.0003
0.0002
Total 0.053267 26
(b) The P-value < 0.0001. There is a significant time effect.
(c) The P-value = 0.0003. There is a significant copper effect.
(d) The interaction effect is significant since the P-value = 0.0002. The interaction
plot is show here.
1
1
1
0.24 0.26 0.28 0.30 0.32 0.34 0.36
Copper Content
Algae Concentration
2
2
2
3
3
3
1 2 3
Time
123
5
12
18
The algae concentrations for the various copper contents are all clearly influenced
by the time effect shown in the graph.
14.13 (a) The interaction plot is show here. There seems no interaction effect.
1
1
1.90 1.95 2.00 2.05 2.10 2.15
Treatment
Magnesium Uptake
2
2
1 2
Time
12
12
222 Chapter 14 Factorial Experiments (Two or More Factors)
(b) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Treatment
Time
Treatment*Time
Error
0.060208
0.060208
0.000008
0.003067
1
1
1
8
0.060208
0.060208
0.000008
0.000383
157.07
157.07
0.02
< 0.0001
< 0.0001
0.8864
Total 0.123492 11
(c) The magnesium uptake are lower using treatment 2 than using treatment 1, no
matter what the times are. Also, time 2 has lower magnesium uptake than time
1. All the main effects are significant.
(d) Using the regression model and making “Treatment” as categorical, we have the
following fitted model:
ˆy = 2.4433 − 0.13667 Treatment − 0.13667 Time − 0.00333Treatment × Time.
(e) The P-value of the interaction for the above regression model is 0.8864 and hence
it is insignificant.
14.14 (a) A natural linear model with interaction would be
y = β0 + β1x1 + β2x2 + β12x1x2.
The fitted model would be
ˆy = 0.41772 − 0.06631x1 − 0.00866x2 + 0.00416x1x2,
with the P-values of the t-tests on each of the coefficients as 0.0092, 0.0379 and
0.0318 for x1, x2, and x1x2, respectively. They are all significant at a level larger
than 0.0379. Furthermore, R2
adj = 0.1788.
(b) The new fitted model is
ˆy = 0.3368 − 0.15965x1 + 0.02684x2 + 0.00416x1x2 + 0.02333x2
1 − 0.00155x2
2,
with P-values of the t-tests on each of the coefficients as 0.0004, < 0.0001, 0.0003,
0.0156, and < 0.0001 for x1, x2. x1x2, x2
1, and x2
2, respectively. Furthermore,
R2
adj = 0.7700 which is much higher than that of the model in (a). Model in (b)
would be more appropriate.
14.15 The ANOVA table is given here.
Solutions for Exercises in Chapter 14 223
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Main effect
A
B
C
Two-factor
Interaction
AB
AC
BC
Three-factor
Interaction
ABC
Error
2.24074
56.31815
17.65148
31.47148
31.20259
2156074
26.79852
148.04000
1
2
2
2
2
4
4
36
2.24074
28.15907
8.82574
15.73574
15.60130
5.39019
6.69963
4.11221
0.54
6.85
3.83
3.83
3.79
1.31
1.63
0.4652
0.0030
0.1316
0.0311
0.0320
0.2845
0.1881
Total 335.28370 53
(a) Based on the P-values, only AB and AC interactions are significant.
(b) The main effect B is significant. However, due to significant interactions mentioned
in (a), the insignificance of A and C cannot be counted.
(c) Look at the interaction plot of the mean responses versus C for different cases of
A.
1
1
1
13.5 14.0 14.5 15.0 15.5 16.0
C
y
2
2
2
1 2 3
A
12
12
Apparently, the mean responses at different levels of C varies in different patterns
for the different levels of A. Hence, although the overall test on factor C is
insignificant, it is misleading since the significance of the effect C is masked by
the significant interaction between A and C.
14.16 (a) When only A, B, C, and BC factors are in the model, the P-value for BC
interaction is 0.0806. Hence at level of 0.05, the interaction is insignificant.
(b) When the sum of squares of the BC term is pooled with the sum of squares of
the error, we increase the degrees of freedom of the error term. The P-values of
224 Chapter 14 Factorial Experiments (Two or More Factors)
the main effects of A, B, and C are 0.0275, 0.0224, and 0.0131, respectively. All
these are significant.
14.17 Letting A, B, and C designate coating, humidity, and stress, respectively, the ANOVA
table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Main effect
A
B
C
Two-factor
Interaction
AB
AC
BC
Three-factor
Interaction
ABC
Error
216, 384.1
19, 876, 891.0
427, 993, 946.4
31, 736, 625
699, 830.1
58, 623, 693.2
36, 034, 808.9
335, 213, 133.6
1
2
2
2
2
4
4
72
216, 384.1
9, 938, 445.5
213, 996, 973.2
15, 868, 312.9
349, 915.0
13, 655, 923.3
9, 008, 702.2
4, 655, 738.0
0.05
2.13
45.96
3.41
0.08
3.15
1.93
0.8299
0.1257
< 0.0001
0.0385
0.9277
0.0192
0.1138
Total 910, 395, 313.1 89
(a) The Coating and Humidity interaction, and the Humidity and Stress interaction
have the P-values of 0.0385 and 0.0192, respectively. Hence, they are all significant.
On the other hand, the Stress main effect is strongly significant as well.
However, both other main effects, Coating and Humidity, cannot be claimed as
insignificant, since they are all part of the two significant interactions.
(b) A Stress level of 20 consistently produces low fatigue. It appears to work best
with medium humidity and an uncoated surface.
14.18 The ANOVA table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
A
B
C
AB
AC
BC
ABC
Error
1.90591
0.02210
38.93402
0.88632
0.53594
0.45435
0.42421
2.45460
3
1
1
3
3
1
3
32
0.63530
0.02212
38.93402
0.29544
0.17865
0.45435
0.14140
0.07671
8.28
0.29
507.57
3.85
2.33
5.92
1.84
0.0003
0.5951
< 0.0001
0.0185
0.0931
0.0207
0.1592
Total 45.61745 47
Solutions for Exercises in Chapter 14 225
(a) Two-way interactions of AB and BC are all significant and main effect of A and
C are all significant. The insignificance of the main effect B may not be valid due
to the significant BC interaction.
(b) Based on the P-values, Duncan’s tests and the interaction means, the most important
factor is C and using C = 2 is the most important way to increase percent
weight. Also, using factor A at level 1 is the best.
14.19 The ANOVA table shows:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
A
B
C
AB
AC
BC
ABC
Error
0.16617
0.07825
0.01947
0.12845
0.06280
0.12644
0.14224
0.47323
2
2
2
4
4
4
8
81
0.08308
0.03913
0.00973
0.03211
0.01570
0.03161
0.01765
0.00584
14.22
6.70
1.67
5.50
2.69
5.41
3.02
< 0.0001
0.0020
0.1954
0.0006
0.0369
0.0007
0.0051
Total 1.19603 107
There is a significant three-way interaction by Temperature, Surface, and Hrc. A plot
of each Temperature is given to illustrate the interaction
1
1
1
0.55 0.60 0.65 0.70 0.75
Temperature=Low
Hrc
Gluing Power
2
2
2
3
3
3
20 40 60
Surface
123
123
1
1 1
0.45 0.50 0.55 0.60 0.65
Temperature=Medium
Hrc
Gluing Power 2
2
2
3
3
3
20 40 60
Surface
123
123
1
1
1
0.40 0.45 0.50 0.55 0.60
Temperature=High
Hrc
Gluing Power
2
2
2
3
3
3
20 40 60
Surface
123
123
226 Chapter 14 Factorial Experiments (Two or More Factors)
14.20 (a) yijk = μ + αi + βj + γk + (βγ)jk + ǫijk;
P
j
βj = 0,
P
k
γk = 0,
P
j
(βγ)jk = 0,
P
k
(βγ)jk = 0, and ǫijk ∼ n(x; 0, σ2).
(b) The P-value of the Method and Type of Gold interaction is 0.10587. Hence, the
interaction is at least marginally significant.
(c) The best method depends on the type of gold used.
The tests of the method effect for different type of gold yields the P-values as
0.9801 and 0.0099 for “Gold Foil” and “Goldent”, respectively. Hence, the methods
are significantly different for the “Goldent” type.
Here is an interaction plot.
1 1
1
600 650 700 750 800
Method
Hardness
2
2
2
1 2 3
Type
12
foil
goldent
It appears that when Type is “Goldent” and Method is 1, it yields the best
hardness.
14.21 (a) Yes, the P-values for Brand ∗ Type and Brand ∗ Temp are both < 0.0001.
(b) The main effect of Brand has a P-value < 0.0001. So, three brands averaged
across the other two factore are significantly different.
(c) Using brand Y , powdered detergent and hot water yields the highest percent
removal of dirt.
14.22 (a) Define A, B, and C as “Powder Temperature,” “Die Temperature,” and “Extrusion
Rate,” respectively. The ANOVA table shows:
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
A
B
C
AB
AC
Error
78.125
3570.125
2211.125
0.125
1.125
1.25
1
1
1
1
1
2
78.125
3570.125
2211.125
0.125
1.125
0.625
125.00
5712.20
3537.80
0.20
1.80
0.0079
0.0002
0.0003
0.6985
0.3118
Total 5861.875 7
Solutions for Exercises in Chapter 14 227
The ANOVA results only show that the main effects are all significant and no
two-way interactions are significant.
(b) The interaction plots are shown here.
1
1
100 110 120 130 140
Powder Temperature
Radius
2
2
150 190
DieTemp
12
220
250
1
1
110 120 130 140
Powder Temperature
Radius
2
2
150 190
Rate
12
12
24
(c) The interaction plots in part (b) are consistent with the findings in part (a) that
no two-way interactions present.
14.23 (a) The P-values of two-way interactions Time×Temperature, Time×Solvent,
Temperature × Solvent, and the P-value of the three-way interaction
Time×Temperature×Solvent are 0.1103, 0.1723, 0.8558, and 0.0140, respectively.
(b) The interaction plots for different levels of Solvent are given here.
1 1
1
91 92 93 94 95
Solvent=Ethanol
Time
Amount of Gel
2 2
2
4 8 16
Temp
12
80
120
1
1
1 91 92 93 94 95
Solvent=Toluene
Time
Amount of Gel
2
2
2
4 8 16
Temp
12
80
120
(c) A normal probability plot of the residuals is given and it appears that normality
assumption may not be valid.
−2 −1 0 1 2
−0.3 −0.2 −0.1 0.0 0.1 0.2 0.3
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
228 Chapter 14 Factorial Experiments (Two or More Factors)
14.24 (a) The two-way interaction plots are given here and they all show significant interactions.
1
1
1
0.50 0.55 0.60 0.65
Surface
Power
2
2
2
3 3
3
1 2 3
Temp
123
123
1
1
1
0.50 0.55 0.60 0.65
Hardness
Power
2 2
2
3
3
3
20 40 60
Temp
123
123
1
1
1
0.50 0.54 0.58 0.62
Hardness
Power
2
2
2
3
3
3
20 40 60
Surface
123
123
(b) The normal probability plot of the residuals is shown here and it appears that
normality is somewhat violated at the tails of the distribution.
−2 −1 0 1 2
−0.1 0.0 0.1 0.2
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
14.25 The ANOVA table is given.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Filters
Operators
Interaction
Error
4.63389
10.31778
1.65722
4.44000
2
3
6
24
2.31694
3.43926
0.27620
0.18500
8.39
12.45
1.49
0.0183
0.0055
0.2229
Total 21.04889 35
Solutions for Exercises in Chapter 14 229
Note the f values for the main effects are using the interaction term as the denominator.
(a) The hypotheses are
H0 : σ2
= 0,
H1 : σ2
6= 0.
Decision: Since P-value = 0.2229, the null hypothesis cannot be rejected. There
is no significant interaction variance component.
(b) The hypotheses are
H
′
0 : σ2
= 0. H
′′
0 : σ2
= 0.
H
′
1 : σ2
6= 0. H
′′
1 : σ2
6= 0.
Decisions: Based on the P-values of 0.0183, and 0.0055 for H′
0 and H′
1, respectively,
we reject both H′
0 and H′′
0 . Both σ2
and σ2
are significantly different from
zero.
(c) ˆσ2 = s2 = 0.185; ˆσ2
= 2.31691−0.185
12 = 0.17766, and ˆσ2
= 3.43926−0.185
9 = 0.35158.
14.26 ˆσ2
= (42.6289/2−14.8011/4
12 = 1.4678 Brand.
ˆσ2
= (299.3422/2−14.8011/4
12 = 12.1642 Time.
s2 = 0.9237.
14.27 The ANOVA table with expected mean squares is given here.
Source of Degrees of Mean (a) (b)
Variation Freedom Square Computed f Computed f
A
B
C
AB
AC
BC
ABC
Error
3
1
2
3
6
2
6
24
140
480
325
15
24
18
2
5
f1 = s2
1 /s2
5 = 5.83
f2 = s2
2 /s2
p1 = 78.82
f3 = s2
3 /s2
5 = 13.54
f4 = s2
4 /s2
p2 = 2.86
f5 = s2
5 /s2
p2 = 4.57
f6 = s2
6 /s2
p1 = 4.09
f7 = s2
7 /s2 = 0.40
f1 = s2
1 /s2
5 = 5.83
f2 = s2
2 /s2
6 = 26.67
f3 = s2
3 /s2
5 = 13.54
f4 = s2
4 /s2
7 = 7.50
f5 = s2
5 /s2
7 = 12.00
f6 = s2
6 /s2
7 = 9.00
f7 = s2
7 /s2 = 0.40
Total 47
In column (a) we have found the following main effects and interaction effects significant
using the pooled estimates: σ2
, σ2
, and σ2
.
s2
p1 = (12 + 120)/30 = 4.4 with 30 degrees of freedom.
s2
p2 = (12 + 120 + 36)/32 = 5.25 with 32 degrees of freedom.
s2
p3 = (12 + 120 + 36 + 45)/35 = 6.09 with 35 degrees of freedom.
In column (b) we have found the following main effect and interaction effect significant
when sums of squares of insignificant effects were not pooled: σ2
and σ2
.
230 Chapter 14 Factorial Experiments (Two or More Factors)
14.28
P4
i=1
γ2
k = 0.24 and φ =
q
(16)(0.24)
(3)(0.197) = 2.55. With α = 0.05, v1 = 2 and v2 = 39 we find
from A.16 that the power is approximately 0.975. Therefore, 2 observations for each
treatment combination are sufficient.
14.29 The power can be calculated as
1 − β = P
"
F(2, 6) > f0.05(2, 6)
σ2 + 3σ2
σ2 + 3σ2
+ 12σ2
#
= P
F(2, 6) >
(5.14)(0.2762)
2.3169
= P[F(2, 6) > 0.6127] = 0.57.
14.30 (a) A mixed model. Inspectors (αi in the model) are random effects. Inspection level
(βj in the model) is a fixed effect.
yijk = μ + αi + βj + (αβ)ij + ǫijk;
αi ∼ n(x; 0, σ2
), (αβ)ij ∼ n(x; 0, σ2
), ǫijk ∼ n(x; 0, σ2),
X
j
βj = 0.
(b) The hypotheses are
H0 : σ2
= 0. H0 : σ2
= 0. H0 : β1 = β2 = β3 = 0,
H1 : σ2
6= 0. H1 : σ2
6= 0 H1 : At least one βi’s is not 0.
f2,36 = 0.02 with P-value = 0.9806. There does not appear to be an effect due to
the inspector.
f4,36 = 0.04 with P-value = 0.9973. There does not appear to be an effect due to
the inspector by inspector level.
f2,4 = 54.77 with P-value = 0.0012. Mean inspection levels were significantly
different in determining failures per 1000 pieces.
14.31 (a) A mixed model.
(b) The ANOVA table is
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Material
Brand
Material*Brand
Error
1.03488
0.60654
9, 70109
0.09820
2
2
4
9
0.51744
0.30327
0.17527
0.01091
47.42
1.73
16.06
< 0.0001
0.2875
0.0004
Total 2.44071 17
(c) No, the main effect of Brand is not significant. An interaction plot is given.
Solutions for Exercises in Chapter 14 231
1
1
1
4.6 4.8 5.0 5.2 5.4 5.6
Brand
Year
2 2
3 2
3
3
A B C
Material
123
AB
C
Although brand A has highest means in general, it is not always significant,
especially for Material 2.
14.32 (a) Operators (αi) and time of day (βj) are random effects.
yijk = μ + αi + βj + (αβ)ij + ǫijk;
αi ∼ n(x; 0, σ2
), βj ∼ n(x; 0, σ2
), (αβ)ij ∼ n(x; 0, σ2
), ǫijk ∼ n(x; 0, σ2).
(b) σ2
= σ2
= 0 (both estimates of the variance components were negative).
(c) The yield does not appear to depend on operator or time.
14.33 (a) A mixed model. Power setting (αi in the model) is a fixed effect. Cereal type (βj
in the model) is a random effect.
yijk = μ + αi + βj + (αβ)ij + ǫijk;
X
i
αi = 0, βj ∼ n(x; 0, σ2
), (αβ)ij ∼ n(x; 0, σ2
), ǫijk ∼ n(x; 0, σ2).
(b) No. f2,4 = 1.37 and P-value = 0.3524.
(c) No. The estimate of σ2
is negative.
14.34 (a) The ANOVA table is given.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Sweetener
Flour
Interaction
Error
0.00871
0.00184
0.01015
0.01600
3
1
3
16
0.00290
0.00184
0.00338
0.00100
2.90
1.84
3.38
0.0670
0.1941
0.0442
Total 0.03670 23
Sweetener factor is close to be significant, while the P-value of the Flour shows
insignificance. However, the interaction effects appear to be significant.
232 Chapter 14 Factorial Experiments (Two or More Factors)
(b) Since the interaction is significant with a P-value = 0.0442, testing the effect
of sweetener on the specific gravity of the cake samples by flour type we get
P-value = 0.0077 for “All Purpose” flour and P-value = 0.6059 for “Cake” flour.
We also have the interaction plot which shows that sweetener at 100% concentration
yielded a specific gravity significantly lower than the other concentrations
for all-purpose flour. For cake flour, however, there were no big differences in the
effect of sweetener concentration.
1
1
1
1
0.80 0.82 0.84 0.86 0.88
Sweetener
Attribute
2
2
2
2
0 50 75 100
Flour
12
all−pupose
cake
14.35 (a) The ANOVA table is given.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Sauce
Fish
Sauce*Fish
Error
1, 031.3603
16, 505.8640
724.6107
3, 381.1480
1
2
2
24
1, 031.3603
8, 252.9320
362.3053
140.8812
7.32
58.58
2.57
0.0123
< 0.0001
0.0973
Total 21, 642.9830 29
Interaction effect is not significant.
(b) Both P-values of Sauce and Fish Type are all small enough to call significance.
14.36 (a) The ANOVA table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Plastic Type
Humidity
Interaction
Error
142.6533
143.7413
133.9400
50.5950
2
3
6
12
71.3267
47.9138
22.3233
4.2163
16.92
11.36
5.29
0.0003
0.0008
0.0070
Total 470.9296 23
The interaction is significant.
(b) The SS for AB with only Plastic Type A and B is 24.8900 with 3 degrees of
freedom. Hence f = 24.8900/3
4.2163 = 1.97 with P-value = 0.1727. Hence, there is no
significant interaction when only A and B are considered.
Solutions for Exercises in Chapter 14 233
(c) The SS for the single-degree-of-freedom contrast is 143.0868. Hence f = 33.94
with P-value < 0.0001. Therefore, the contrast is significant.
(d) The SS for Humidity when only C is considered in Plastic Type is 2.10042. So,
f = 0.50 with P-value = 0.4938. Hence, Humidity effect is insignificant when
Type C is used.
14.37 (a) The ANOVA table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Environment
Stress
Interaction
Error
0.8624
40.8140
0.0326
0.6785
1
2
2
12
0.8624
20.4020
0.0163
0.0565
15.25
360.94
0.29
0.0021
< 0.0001
0.7547
Total 42.3875 17
The interaction is insignificant.
(b) The mean fatigue life for the two main effects are all significant.
14.38 The ANOVA table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Sweetener
Flour
Interaction
Error
1.26893
1.77127
0.14647
2.55547
3
1
3
16
0.42298
1.77127
0.04882
0.15972
2.65
11.09
0.31
0.0843
0.0042
0.8209
Total 5.74213 23
The interaction effect is insignificant. The main effect of Sweetener is somewhat insignificant,
since the P-value = 0.0843. The main effect of Flour is strongly significant.
14.39 The ANOVA table is given here.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
A
B
C
AB
AC
Error
1133.5926
26896.2593
40.1482
216.5185
1.6296
2.2963
2.5926
1844.0000
2
2
2
4
4
4
8
27
566.7963
13448.1296
20.0741
54.1296
0.4074
0.5741
0.3241
68.2963
8.30
196.91
0.29
0.79
0.01
0.01
0.00
0.0016
< 0.0001
0.7477
0.5403
0.9999
0.9999
1.0000
Total 30137.0370 53
234 Chapter 14 Factorial Experiments (Two or More Factors)
All the two-way and three-way interactions are insignificant. In the main effects, only
A and B are significant.
14.40 (a) Treating Solvent as a class variable and Temperature and Time as continuous
variable, only three terms in the ANOVA model show significance. They are (1)
Intercept; (2) Coefficient for Temperature and (3) Coefficient for Time.
(b) Due to the factor that none of the interactions are significant, we can claim that
the models for ethanol and toluene are equivalent apart from the intercept.
(c) The three-way interaction in Exercise 14.23 was significant. However, the general
patterns of the gel generated are pretty similar for the two Solvent levels.
14.41 The ANOVA table is displayed.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Surface
Pressure
Interaction
Error
2.22111
39.10778
112.62222
565.72000
2
2
4
9
1.11056
19.55389
28.15556
62.85778
0.02
0.31
0.45
0.9825
0.7402
0.7718
Total 719.67111 17
All effects are insignificant.
14.42 (a) This is a two-factor fixed-effects model with interaction.
yijk = μ + αi + βj+)αβ)ij + ǫijk,
X
i
αi = 0,
X
j
βj = 0,
X
i
(αβ)ij =
X
j
(αβ)ij = 0, ǫijk ∼ n(x; 0, σ)
(b) The ANOVA table is displayed.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Time
Temperature
Interaction
Error
0.16668
0.27151
0.03209
0.01370
3
2
6
12
0.05556
0.13575
0.00535
0.00114
48.67
118.91
4.68
< 0.0001
< 0.0001
0.0111
Total 0.48398 23
The interaction is insignificant, while two main effects are significant.
(c) It appears that using a temperature of −20◦C with drying time of 2 hours would
speed up the process and still yield a flavorful coffee. It might be useful to try
some additional runs at this combination.
Solutions for Exercises in Chapter 14 235
14.43 (a) Since it is more reasonable to assume the data come from Poisson distribution,
it would be dangerous to use standard analysis of variance because the normality
assumption would be violated. It would be better to transform the data to get at
least stable variance.
(b) The ANOVA table is displayed.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Teller
Time
Interaction
Error
40.45833
97.33333
8.66667
25.50000
3
2
6
12
13.48611
48.66667
1.44444
2.12500
6.35
22.90
0.68
0.0080
< 0.0001
0.6694
Total 171.95833 23
The interaction effect is insignificant. Two main effects, Teller and Time, are all
significant.
(c) The ANOVA table using a squared-root transformation on the response is given.
Source of Sum of Degrees of Mean Computed
Variation Squares Freedom Square f P-value
Teller
Time
Interaction
Error
0.05254
1.32190
0.11502
0.35876
3
2
6
12
0.17514
0.66095
0.01917
0.02990
5.86
22.11
0.64
0.0106
< 0.0001
0.6965
Total 2.32110 23
Same conclusions as in (b) can be reached. To check on whether the assumption
of the standard analysis of variance is violated, residual analysis may used to do
diagnostics.
Chapter 15
2k Factorial Experiments and
Fractions
15.1 Either using Table 15.5 (e.g., SSA = (−41+51−57−63+67+54−76+73)2
24 = 2.6667) or running
an analysis of variance, we can get the Sums of Squares for all the factorial effects.
SSA = 2.6667, SSB = 170.6667, SSC = 104.1667, SS(AB) = 1.500.
SS(AC) = 42.6667, SS(BC) = 0.0000, SS(ABC) = 1.5000.
15.2 A simplified ANOVA table is given.
Source of Degrees of Computed
Variation Freedom f P-value
A
B
AB
C
AC
BC
ABC
Error
1
1
1
1
1
1
1
8
1294.65
43.56
20.88
116.49
16.21
0.00
289.23
< 0.0001
0.0002
0.0018
< 0.0001
0.0038
0.9668
< 0.0001
Total 15
All the main and interaction effects are significant, other than BC effect. However,
due to the significance of the 3-way interaction, the insignificance of BC effect cannot
be counted. Interaction plots are given.
237
238 Chapter 15 2k Factorial Experiments and Fractions
1
1
5 10 15 20
C=−1
A
y
2
2
−1 1
B
12
−1
1
1
1
6 8 10 12 14 16 18
C=1
A
y
2
2
−1 1
B
12
−1
1
15.3 The AD and BC interaction plots are printed here. The AD plot varies with levels of
C since the ACD interaction is significant, or with levels of B since ABD interaction
is significant.
1
1
26.5 27.0 27.5 28.0 28.5 29.0
AD Interaction
D
y
2
2
−1 1
A
12
−1
1
1
1
26.0 27.0 28.0 29.0
BC Interaction
C
y
2
2
−1 1
B
12
−1
1
15.4 The ANOVA table is displayed.
Source of Degrees of Computed
Variation Freedom f P-value
A
B
AB
C
AC
BC
ABC
D
AD
BD
ABD
CD
ACD
BCD
ABCD
Error
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
16
57.85
7.52
6.94
127.86
7.08
10.96
1.26
44.72
4.85
4.85
1.14
6.52
1.72
1.20
0.87
< 0.0001
0.0145
0.0180
< 0.0001
0.0171
0.0044
0.2787
< 0.0001
0.0427
0.0427
0.3017
0.0213
0.2085
0.2900
0.3651
Total 31
Solutions for Exercises in Chapter 15 239
All main effects and two-way interactions are significant, while all higher order interactions
are insignificant.
15.5 The ANOVA table is displayed.
Source of Degrees of Computed
Variation Freedom f P-value
A
B
C
D
AB
AC
AD
BC
BD
CD
Error
1
1
1
1
1
1
1
1
1
1
5
9.98
0.20
6.54
0.02
1.83
0.20
0.57
19.03
1.83
0.02
0.0251
0.6707
0.0508
0.8863
0.2338
0.6707
0.4859
0.0073
0.2338
0.8863
Total 15
One two-factor interaction BC, which is the interaction of Blade Speed and Condition
of Nitrogen, is significant. As of the main effects, Mixing time (A) and Nitrogen
Condition (C) are significant. Since BC is significant, the insignificant main effect B,
the Blade Speed, cannot be declared insignificant. Interaction plots for BC at different
levels of A are given here.
1
1
15.9 16.1 16.3 16.5
Speed and Nitrogen Interaction (Time=1)
Nitrogen Condition
y
2
2
1 2
Speed
12
1 1
15.6 15.7 15.8 15.9 16.0
Speed and Nitrogen Interaction (Time=2)
Nitrogen Condition
y
2
2
1 2
Speed
12
12
15.6 (a) The three effects are given as
wA =
301 + 304 − 269 − 292
4
= 11, wB =
301 + 269 − 304 − 292
4
= −6.5,
wAB =
301 − 304 − 269 + 292
4
= 5.
There are no clear interpretation at this time.
(b) The ANOVA table is displayed.
240 Chapter 15 2k Factorial Experiments and Fractions
Source of Degrees of Computed
Variation Freedom f P-value
Concentration
Feed Rate
Interaction
Error
1
1
1
4
35.85
12.52
7.41
0.0039
0.0241
0.0529
Total 7
The interaction between the Feed Rate and Concentration is closed to be significant
at 0.0529 level. An interaction plot is given here.
1
1
135 140 145 150
Feed Rate
y
2
2
−1 1
Concentration
12
−1
1
The mean viscosity does not change much at high level of concentration, while it
changes a lot at low concentration.
(c) Both main effects are significant. Averaged across Feed Rate a high concentration
of reagent yields significantly higher viscosity, and averaged across concentration
a low level of Feed Rate yields a higher level of viscosity.
15.7 Both AD and BC interaction plots are shown in Exercise 15.3. Here is the interaction
plot of AB.
1
1
26.5 27.0 27.5 28.0 28.5
AB Interaction
A
y
2
2
−1 1
B
12
−1
1
For AD, at the high level of A, Factor D essentially has no effect, but at the low level
of A, D has a strong positive effect. For BC, at the low level of B, Factor C has a
strong negative effect, but at the high level of B, the negative effect of C is not as
pronounced. For AB, at the high level of B, A clearly has no effect. At the low level
of B, A has a strong negative effect.
15.8 The two interaction plots are displayed.
Solutions for Exercises in Chapter 15 241
1
1
26 27 28 29 30
AD Interaction (B=−1)
A
y
2
2
−1 1
D
12
−1
1
1
1
25.0 25.5 26.0 26.5 27.0 27.5
AD Interaction (B=1)
A
y
2
2
−1 1
D
12
−1
1
It can be argued that when B = 1 that there is essentially no interaction between A
and D. Clearly when B = −1, the presence of a high level of D produces a strong
negative effect of Factor A on the response.
15.9 (a) The parameter estimates for x1, x2 and x1x2 are given as follows.
Variable Degrees of Freedom Estimate f P-value
x1
x2
x1x2
1
1
1
5.50
−3.25
2.50
5.99
−3.54
2.72
0.0039
0.0241
0.0529
(b) The coefficients of b1, b2, and b12 are wA/2, wB/2, and wAB/2, respectively.
(c) The P-values are matched exactly.
15.10 The effects are given here.
A B C D AB AC AD BC
−0.2625 −0.0375 0.2125 0.0125 −0.1125 0.0375 −0.0625 0.3625
BD CD ABC ABD ACD BCD ABCD
0.1125 0.0125 −0.1125 0.0375 −0.0625 0.1125 −0.0625
The normal probability plot of the effects is displayed.
−1 0 1
−0.2 0.0 0.1 0.2 0.3
Normal Q−Q Plot
Theoretical Quantiles
Sample Quantiles
A
AB ABC
AD ACD ABCD
B
D CD
AC ABD
BD BCD
C
BC
(a) It appears that all three- and four-factor interactions are not significant.
(b) From the plot, it appears that A and BC are significant and C is somewhat
significant.
242 Chapter 15 2k Factorial Experiments and Fractions
15.11 (a) The effects are given here and it appears that B, C, and AC are all important.
A B C AB AC BC ABC
−0.875 5.875 9.625 −3.375 −9.625 0.125 −1.125
(b) The ANOVA table is given.
Source of Degrees of Computed
Variation Freedom f P-value
A
B
AB
C
AC
BC
ABC
Error
1
1
1
1
1
1
1
8
0.11
4.79
12.86
1.58
12.86
0.00
0.18
0.7528
0.0600
0.0071
0.2440
0.0071
0.9640
0.6861
Total 15
1
1
30 35 40 45 50
AC Interaction
A
y
2
2
−1 1
C
12
−1
1
(c) Yes, they do agree.
(d) For the low level of Cutting Angle, C, Cutting Speed, A, has a positive effect on
the life of a machine tool. When the Cutting Angle is large, Cutting Speed has a
negative effect.
15.12 A is not orthogonal to BC, B is not orthogonal to AC, and C is not orthogonal to AB.
If we assume that interactions are negligible, we may use this experiment to estimate
the main effects. Using the data, the effects can be obtained as A: 1.5; B: −6.5;
C: 2.5. Hence Factor B, Tool Geometry, seems more significant than the other two
factors.
15.13 Here is the block arrangement.
Block Block Block
1 2 1 2 1 2
(1)
c
ab
abc
a
b
ac
bc
(1)
c
ab
abc
a
b
ac
bc
(1)
c
ab
abc
a
b
ac
bc
Replicate 1 Replicate 2 Replicate 3
AB Confounded AB Confounded AB Confounded
Solutions for Exercises in Chapter 15 243
Analysis of Variance
Source of Variation Degrees of Freedom
Blocks
A
B
C
AC
BC
ABC
Error
5
1
1
1
1
1
1
12
Total 23
15.14 (a) ABC is confounded with blocks in the first replication and ABCD is confounded
with blocks in second replication.
(b) Computing the sums of squares by the contrast method yields the following
ANOVA table.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
Blocks
A
B
C
D
AB
AC
BC
AD
BD
CD
ABC
ABD
ACD
BCD
ABCD
Error
3
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
13
2.32
2.00
0.50
4.50
8.00
0.50
0.32
0.50
0.72
0.32
0.18
1.16
0.32
0.02
0.18
0.53
0.60
3.34
0.83
7.51
13.36
0.83
0.53
0.83
1.20
0.53
0.30
1.93
0.53
0.03
0.30
0.88
0.0907
0.3775
0.0168
0.0029
0.3775
0.4778
0.3775
0.2928
0.4778
0.5928
0.1882
0.4778
0.8578
0.5928
0.3659
Total 31
Only the main effects C and D are significant.
15.15 L1 = γ1 + γ2 + γ3 and L2 = γ1 + γ2 + γ4. For treatment combination (1) we find
L1 (mod 2) = 0. For treatment combination a we find L1 (mod 2) = 1 and L2 (mod 2) =
1. After evaluating L1 and L2 for all 16 treatment combinations we obtain the following
blocking scheme:
244 Chapter 15 2k Factorial Experiments and Fractions
Block 1 Block 2 Block 3 Block 4
(1)
ab
acd
bcd
c
abc
ad
bd
d
ac
bc
abd
a
b
cd
abcd
L1 = 0 L1 = 1 L1 = 0 L1 = 1
L2 = 0 L2 = 0 L2 = 1 L2 = 1
Since (ABC)(ABD) = A2B2CD = CD (mod 2), then CD is the other effect confounded.
15.16 (a) L1 = γ1+γ2+γ4 +γ5, L2 = γ1 +γ5. We find that the following treatment combinations
are in the principal block (L1 = 0, L2 = 0): (1), c, ae, bd, ace, abde, abcde.
The other blocks are constructed by multiplying the treatment combinations in
the principal block modulo 2 by a, b, and ab, respectively, to give the following
blocking arrangement:
Block 1 Block 2 Block 3 Block 4
(1)
c
ae
bd
ace
bcd
abde
abcde
a
ac
e
abd
ce
abcd
bde
bcde
b
bc
abe
d
abce
cd
ade
acde
ab
abc
bce
ad
bce
acd
de
cde
(b) (ABDE)(AE) = BD (mod 2). Therefore BD is also confounded with days.
(c) Yates’ technique gives the following sums of squares for the main effects:
SSA = 21.9453, SSB = 40.2753, SSC = 2.4753,
SSD = 7.7028, SSE = 1.0878.
15.17 L1 = γ1 + γ2 + γ3, L2 = γ1 + γ2.
Block Block Block
1 2 1 2 1 2
abc
a
b
c
ab
ac
bc
(1)
abc
a
b
c
ab
ac
bc
(1)
(1)
c
ab
abc
a
b
ac
bc
Rep 1 Rep 2 Rep 3
ABC Confounded ABC Confounded AB Confounded
Solutions for Exercises in Chapter 15 245
For treatment combination (1) we find L1 (mod 2) = 0 and L2 (mod 2) = 0. For
treatment combination a we find L1 (mod 2) = 1 and L2 (mod 2) = 1. Replicate 1 and
Replicate 2 have L1 = 0 in one block and L1 = 1 in the other. Replicate 3 has L2 = 0
in one block and L2 = 1 in the other.
Analysis of Variance
Source of Variation Degrees of Freedom
Blocks
A
B
C
AB
AC
BC
ABC
Error
5
1
1
1
1′
1
1
1′
11
Total 23
Relative information on ABC = 1
3 and relative information on AB = 2
3 .
15.18 (a) The ANOVA table is shown here.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
Operators
A
B
C
D
Error
1
1
1
1
1
10
0.1225
4.4100
3.6100
9.9225
2.2500
2.8423
0.04
1.55
1.27
3.49
0.79
0.2413
0.2861
0.0912
0.3945
Total 15
None of the main effects is significant at 0.05 level.
(b) ABC is confounded with operators since all treatments with positive signs in the
ABC contrast are in one block and those with negative signs are in the other
block.
15.19 (a) One possible design would be:
Machine
1
2
3
4
(1)
a
c
d
ab
b
abc
abd
ce
ace
e
cde
abce
bce
abe
abcde
acd
cd
ad
ac
bde
abde
bcde
be
ade
de
acde
ae
bcd
abcd
bd
bc
(b) ABD, CDE, and ABCE.
246 Chapter 15 2k Factorial Experiments and Fractions
15.20 (a) ˆy = 43.9 + 1.625x1 − 8.625x2 + 0.375x3 + 9.125x1x2 + 0.625x1x3 + 0.875x2x3.
(b) The Lack-of-fit test results in a P-value of 0.0493. There are possible quadratic
terms missing in the model.
15.21 (a) The P-values of the regression coefficients are:
Parameter Intercept x1 x2 x3 x1x2 x1x3 x2x3 x1x2x3
P-value < 0.0001 0.5054 0.0772 0.0570 0.0125 0.0205 0.7984 0.6161
and s2 = 0.57487 with 4 degrees of freedom. So x2, x3, x1x2 and x1x3 are important
in the model.
(b) t = ¯yf−¯yC √s2(1/nf+1/nC)
= 52.075−49.275 √(0.57487)(1/8+1/4)
= 6.0306. Hence the P-value = 0.0038 for
testing quadratic curvature. It is significant.
(c) Need one additional design point different from the original ones.
15.22 (a) No.
(b) It could be as follows.
Machine
1
2
3
4
(1)
a
b
d
ad
e
abe
ade
bc
abc
c
bcd
abce
bce
ace
abcde
acd
cd
abcd
ac
abd
bd
ad
ab
cde
acde
bcde
ce
bde
abde
de
be
ADE, BCD and ABCE are confounded with blocks.
(c) Partial confounding.
15.23 To estimate the quadratic terms, it might be good to add points in the middle of the
edges. Hence (−1, 0), (0,−1), (1, 0), and (0, 1) might be added.
15.24 The alias for each effect is obtained by multiplying each effect by the defining contrast
and reducing the exponents modulo 2.
A ≡CDE, AB ≡BCDE, BD≡ABCE, B ≡ABCDE, AC ≡DE,
BE≡ABCD, C ≡ADE, AD≡CE, ABC≡BDE, D ≡ACE,
AE≡CD, ABD≡BCE, E ≡ACD, BC ≡ABDE, ABE≡BCD,
15.25 (a) With BCD as the defining contrast, we have L = γ2+γ3+γ4. The 1
2 fraction corresponding
to L = 0 (mod 2 is the principal block: {(1), a, bc, abc, bd, abd, cd, acd}.
(b) To obtain 2 blocks for the 1
2 fraction the interaction ABC is confounded using
L = γ1 + γ2 + γ3:
Solutions for Exercises in Chapter 15 247
Block 1 Block 2
(1)
bc
abd
acd
a
abc
bd
cd
(c) Using BCD as the defining contrast we have the following aliases:
A≡ABCD, AB≡ACD, B≡CD, AC≡ABD,
C≡BD, AD≡ABC, D≡BC.
Since AD and ABC are confounded with blocks there are only 2 degrees of freedom
for error from the unconfounded interactions.
Analysis of Variance
Source of Variation Degrees of Freedom
Blocks
A
B
C
D
Error
1
1
1
1
1
2
Total 7
15.26 With ABCD and BDEF as defining contrasts, we have
L1 = γ1 + γ2 + γ3 + γ4, L2 = γ2 + γ4 + γ5 + γ6.
The following treatment combinations give L1 = 0, L2 = 0 (mod 2) and thereby suffice
as the 1
4 fraction:
{(1), ac, bd, abcd, abe, bce, ade, abf, bcf, adf, cdf, ef, acef, bdef, abcdef}.
The third defining contrast is given by
(ABCD)(BDEF) = AB2CD2EF = ACEF (mod 2).
The effects that are aliased with the six main effects are:
A≡BCD ≡ABDEF≡CEF, B≡ACD ≡DEF≡ABCEF,
C≡ABD ≡BCDEF≡AEF, D≡ABC ≡BEF ≡ACDEF,
E≡ABCDE≡BDF ≡ACF, F ≡ABCDF≡BDE≡ACE.
248 Chapter 15 2k Factorial Experiments and Fractions
15.27 (a) With ABCE and ABDF, and hence (ABCE)(ABDF) = CDEF as the defining
contrasts, we have
L1 = γ1 + γ2 + γ3 + γ5, L2 = γ1 + γ2 + γ4 + γ6.
The principal block, for which L1 = 0, and L2 = 0, is as follows:
{(1), ab, acd, bcd, ce, abce, ade, bde, acf, bcf, df, abdf, aef, bef, cdef, abcdef}.
(b) The aliases for each effect are obtained by multiplying each effect by the three
defining contrasts and reducing the exponents modulo 2.
A ≡BCE ≡BDF ≡ACDEF, B ≡ACE ≡ADF ≡BCDEF,
C ≡ABE ≡ABCDF≡DEF, D ≡ABCDE≡ABF ≡CEF,
E ≡ABC ≡ABDEF≡CDF, F ≡ABCEF ≡ABD ≡CDE,
AB ≡CE ≡DF ≡ABCDEF, AC ≡BE ≡BCDF≡ADEF,
AD ≡BCDE≡BF ≡ACEF, AE ≡BC ≡BDEF≡ACDF,
AF ≡BCEF ≡BD ≡ACDE, CD ≡ABDE ≡ABCF ≡EF,
DE ≡ABCD≡ABEF ≡CF, BCD≡ADE ≡ACF ≡BEF,
DCF≡AEF ≡ACD ≡BDE, .
Since E and F do not interact and all three-factor and higher interactions are
negligible, we obtain the following ANOVA table:
Source of Variation Degrees of Freedom
A
B
C
D
E
F
AB
AC
AD
BC
BD
CD
Error
1
1
1
1
1
1
1
1
1
1
1
1
3
Total 15
15.28 The ANOVA table is shown here and the error term is computed by pooling all the
interaction effects. Factor E is the only significant effect, at level 0.05, although the
decision on factor G is marginal.
Solutions for Exercises in Chapter 15 249
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
A
B
C
D
E
F
G
Error
1
1
1
1
1
1
1
8
1.44
4.00
9.00
5.76
16.00
3.24
12.96
2.97
0.48
1.35
3.03
1.94
5.39
1.09
4.36
0.5060
0.2793
0.1199
0.2012
0.0488
0.3268
0.0701
Total 15
15.29 All two-factor interactions are aliased with each other. So, assuming that two-factor
as well as higher order interactions are negligible, a test on the main effects is given in
the ANOVA table.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
A
B
C
D
Error
1
1
1
1
3
6.125
0.605
4.805
0.245
1.053
5.81
0.57
4.56
0.23
0.0949
0.5036
0.1223
0.6626
Total 7
Apparently no main effects is significant at level 0.05. Comparatively factors A and C
are more significant than the other two. Note that the degrees of freedom on the error
term is only 3, the test is not very powerful.
15.30 Two-factor interactions are aliased with each other. There are total 7 two-factor interactions
that can be estimated. Among those 7, we picked the three, which are AC,
AF, and BD, that have largest SS values and pool the other 2-way interactions to the
error term. An ANOVA can be obtained.
250 Chapter 15 2k Factorial Experiments and Fractions
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
A
B
C
D
E
F
AC
AF
BD
Error
1
1
1
1
1
1
1
1
1
6
81.54
166.54
5.64
4.41
40.20
1678.54
978.75
625.00
429.53
219.18
0.37
0.76
0.03
0.02
0.18
7.66
4.47
2.85
1.96
0.5643
0.4169
0.8778
0.8918
0.6834
0.0325
0.0790
0.1423
0.2111
Total 15
Main effect F, the location of detection, appears to be the only significant effect. The
AC interaction, which is aliased with BE, has a P-value closed to 0.05.
15.31 To get all main effects and two-way interactions in the model, this is a saturated design,
with no degrees of freedom left for error. Hence, we first get all SS of these effects and
pick the 2-way interactions with large SS values, which are AD, AE, BD and BE.
An ANOVA table is obtained.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
A
B
C
D
E
AD
AE
BD
BE
Error
1
1
1
1
1
1
1
1
1
6
388, 129.00
277, 202.25
4, 692.25
9, 702.25
1, 806.25
1, 406.25
462.25
1, 156.25
961.00
108.25
3, 585.49
2, 560.76
43.35
89.63
16.69
12.99
4.27
10.68
8.88
< 0.0001
< 0.0001
0.0006
< 0.0001
0.0065
0.0113
0.0843
0.0171
0.0247
Total 15
All main effects, plus AD, BD and BE two-way interactions, are significant at 0.05
level.
15.32 Consider a 24 design with letters A, B, C, and D, with design points
{(1), a, b, c, d, ab, ac, ad, bc, bd, cd, abc, abd, acd, bcd, abcd}
. Using E = ABCD, we have the following design:
{e, a, b, c, d, abe, ace, ade, bce, bde, cde, abc, abd, acd, bcd, abcde}.
Solutions for Exercises in Chapter 15 251
15.33 Begin with a 23 with design points
{(1), a, b, c, ab, ac, bc, abc}.
Now, use the generator D = AB, E = AC, and F = BC. We have the following
result:
{def, af, be, cd, abd, ace, bcf, abcdef}.
15.34 We can use the D = AB, E = −AC and F = BC as generators and obtain the result:
{df, aef, b, cde, abde, ac, bcef, abcdf}.
15.35 Here are all the aliases
A≡BD≡CE≡CDF≡BEF ≡ ≡ABCF ≡ADEF ≡ABCDE;
B≡AD≡CF ≡CDE≡AEF ≡ ≡ABCE ≡BDEF ≡ABCDF;
C≡AE≡BF ≡BDE≡ADF ≡ ≡CDEF ≡ABCD ≡ABCEF;
D≡AB≡EF ≡BCE≡ACF ≡ ≡BCDF ≡ACDE ≡ABDEF;
E≡AC ≡DF≡ABF ≡BCD≡ ≡ABDE ≡BCEF ≡ACDEF;
F ≡BC≡DE≡ACD≡ABE≡ ≡ACEF ≡ABDF ≡BCDEF.
15.36 (a) The defining relation is ABC = −I.
(b) A = −BC, B = −AC, and C = −AB.
(c) The mean squares for A, B, and C are 1.50, 0.34, and 5.07, respectively. So,
factor C, the amount of grain refiner, appears to be most important.
(d) Low level of C.
(e) All at the “low” level.
(f) A hazard here is that the interactions may play significant roles. The following
are two interaction plots.
1
1
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
AB Interaction
A
y
2
2
−1 1
B
12
−1
1
1
1
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
BC Interaction
B
y
2
2
−1 1
C
12
−1
1
252 Chapter 15 2k Factorial Experiments and Fractions
15.37 When the variables are centered and scaled, the fitted model is
ˆy = 12.7519 + 4.7194x1 + 0.8656x2 − 1.4156x3.
The lack-of-fit test results in an f-value of 81.58 with P-value < 0.0001. Hence, higherorder
terms are needed in the model.
15.38 The ANOVA table for the regression model looks like the following.
Coefficients Degrees of Freedom
Intercept
β1
β2
β3
β4
β5
Two-factor interactions
Lack of fit
Pure error
1
1
1
1
1
1
10
16
32
Total 63
15.39 The defining contrasts are
AFG, CEFG, ACDF, BEG, BDFG, CDG, BCDE, ABCDEFG, DEF, ADEG.
15.40 Begin with the basic line for N = 24; permute as described in Section 15.12 until 18
columns are formed.
15.41 The fitted model is
ˆy =190, 056.67 + 181, 343.33x1 + 40, 395.00x2 + 16, 133.67x3 + 45, 593.67x4
− 29, 412.33x5 + 8, 405.00x6.
The t-tests are given as
Variable t P-value
Intercept
x1
x2
x3
x4
x5
x6
4.48
4.27
0.95
0.38
1.07
−0.69
0.20
0.0065
0.0079
0.3852
0.7196
0.3321
0.5194
0.8509
Only x1 and x2 are significant.
Solutions for Exercises in Chapter 15 253
15.42 An ANOVA table is obtained.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
Polymer 1
Polymer 2
Polymer 1*Polymer 2
Error
1
1
1
4
172.98
180.50
1.62
0.17
1048.36
1093.94
9.82
< 0.0001
< 0.0001
0.0351
Total 7
All main effects and interactions are significant.
15.43 An ANOVA table is obtained.
Source of Degrees of Mean Computed
Variation Freedom Square f P-value
Mode
Type
Mode*Type
Error
1
1
1
12
2, 054.36
4, 805.96
482.90
27.67
74.25
173.71
17.45
< 0.0001
< 0.0001
0.0013
Total 15
All main effects and interactions are significant.
15.44 Two factors at two levels each can be used with three replications of the experiment,
giving 12 observations. The requirement that there must be tests on main effects
and the interactions suggests that partial confounding be used The following design is
indicated:
Block Block Block
1 2 1 2 1 2
(1)
ab
a
b
a
ab
(1)
b
(1)
a
ab
a
Rep 1 Rep 2 Rep 3
15.45 Using the contrast method and compute sums of squares, we have
Source of Variation d.f. MS f
A
B
C
D
E
F
Error
1
1
1
1
1
1
8
0.0248
0.0322
0.0234
0.0676
0.0028
0.0006
0.0201
1.24
1.61
1.17
3.37
0.14
0.03
254 Chapter 15 2k Factorial Experiments and Fractions
15.46 With the defining contrasts ABCD, CDEFG, and BDF, we have
L1 = γ1 + γ2 + γ3 + γ4,
L2 = γ3 + γ4 + γ5 + γ6 + γ7.
L3 = γ2 + γ4 + γ6.
The principal block and the remaining 7 blocks are given by
Block 1 Block 2 Block 3 Block 4
(1), eg
abcd, bdg
adf, bcf
cdef, abcdeg
bde, adefg
bcefg, cdfg
acg, abef
ace, abfg
a, aeg
bcd, abdg
df, abcf
acdef, bcdeg
abde, defg
abcefg, acdfg
cg, bef
ce, bfg
b, beg
acd, dg
abdf, cf
bcdef, acdeg
de, abdefg
cefg, bcdfg
abcg, aef
abce, afg
c, ceg
abd, bcdg
acdf, bf
def, abdeg
bcde, acdefg
befg, dfg
ag, abcef
ae, abcfg
Block 5 Block 6 Block 7 Block 8
d, deg
abc, bg
af, bcdf
cef, abceg
be, aefg
bcdefg, cfg
acdg, abdef
acde, abdfg
e, g
abcde, bdeg
adef, bcef
cdf, abcdg
bd, adfg
bcfg, cdefg
aceg, abf
ac, abefg
f, efg
abcdf, bdfg
ad, bc
cde, abcdefg
bdef, adeg
bceg, cdg
acfg, abe
acef, abg
ab, abeg
cd, adg
bdf, acf
abcdef, cdeg
ade, bdefg
acefg, abcdfg
bcg, ef
bce, fg
The two-way interactions AB ≡ CD, AC ≡ BD, AD ≡ BC, BD ≡ F, BF ≡ D and
DF ≡ B.
15.47 A design (where L1 = L2 = L3 = L4 = 0 (mod 2) are used) is:
{(1), abcg, abdh, abef, acdf, aceh, adeg, afgh,
bcde, bcfh, bdfg, cdgh, cefg, defh, degh, abcdefgh}
15.48 In the four defining contrasts, BCDE, ACDF, ABCG, and ABDH, the length of
interactions are all 4. Hence, it must be a resolution IV design.
15.49 Assuming three factors the design is a 23 design with 4 center runs.
15.50 (a) Consider a 23−1
III design with ABC ≡ I as defining contrast. Then the design
points are
Solutions for Exercises in Chapter 15 255
x1 x2 x3
−1
1
−1
1
0
0
−1
−1
1
1
0
0
−1
1
1
−1
0
0
For the noncentral design points, ¯x1 = ¯x2 = ¯x3 = 0 and ¯x2
1 = ¯x2
2 = ¯x2
3 = 1. Hence
E(¯yf −¯y0) = β0+β1¯x1+β2¯x2+β3¯x3+β11¯x2
1 +β22¯x2
2 +β33¯x2
3 −β0 = β11+β22+β33.
(b) It is learned that the test for curvature that involves ¯yf − ¯y0 actually is testing
the hypothesis β11 + β22 + β33 = 0.
Chapter 16
Nonparametric Statistics
16.1 The hypotheses
H0 : ˜μ = 20 minutes
H1 : ˜μ > 20 minutes.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: Subtracting 20 from each observation and discarding the zeroes. We
obtain the signs
− + + − + + − + + +
for which n = 10 and x = 7. Therefore, the P-value is
P = P(X ≥ 7 | p = 1/2) =
X10
x=7
b(x; 10, 1/2)
= 1 −
X6
x=0
b(x; 10, 1/2) = 1 − 0.8281 = 0.1719 > 0.05.
Decision: Do not reject H0.
16.2 The hypotheses
H0 : ˜μ = 12
H1 : ˜μ 6= 12.
α = 0.02.
Test statistic: binomial variable X with p = 1/2.
Computations: Replacing each value above and below 12 by the symbol “+” and “−”,
respectively, and discarding the two values which equal to 12. We obtain the sequence
− + − − + + + + − + + − + + − +
257
258 Chapter 16 Nonparametric Statistics
for which n = 16, x = 10 and n/2 = 8. Therefore, the P-value is
P = 2P(X ≥ 10 | p = 1/2) = 2
X16
x=10
b(x; 16, 1/2)
= 2(1 −
X9
x=0
b(x; 16, 1/2)) = 2(1 − 0.7728) = 0.4544 > 0.02.
Decision: Do not reject H0.
16.3 The hypotheses
H0 : ˜μ = 2.5
H1 : ˜μ 6= 2.5.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: Replacing each value above and below 2.5 by the symbol “+” and “−”,
respectively. We obtain the sequence
− − − − − − − + + − + − − − − −
for which n = 16, x = 3. Therefore, μ = np = (16)(0.5) = 8 and σ =
p
(16)(0.5)(0.5) =
2. Hence z = (3.5 − 8)/2 = −2.25, and then
P = 2P(X ≤ 3 | p = 1/2) ≈ 2P(Z < −2.25) = (2)(0.0122) = 0.0244 < 0.05.
Decision: Reject H0.
16.4 The hypotheses
H0 : ˜μ1 = ˜μ2
H1 : ˜μ1 < ˜μ2.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: After replacing each positive difference by a “+” symbol and negative
difference by a “−” symbol, respectively, and discarding the two zero differences, we
have n = 10 and x = 2. Therefore, the P-value is
P = P(X ≤ 2 | p = 1/2) =
X2
x=0
b(x; 10, 1/2) = 0.0547 > 0.05.
Decision: Do not reject H0.
Solutions for Exercises in Chapter 16 259
16.5 The hypotheses
H0 : ˜μ1 − ˜μ2 = 4.5
H1 : ˜μ1 − ˜μ2 < 4.5.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: We have n = 10 and x = 4 plus signs. Therefore, the P-value is
P = P(X ≤ 4 | p = 1/2) =
X4
x=0
b(x; 10, 1/2) = 0.3770 > 0.05.
Decision: Do not reject H0.
16.6 The hypotheses
H0 : ˜μA = ˜μB
H1 : ˜μA 6= ˜μB.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: We have n = 14 and x = 12. Therefore, μ = np = (14)(1/2) = 7 and
σ =
p
(14)(1/2)(1/2) = 1.8708. Hence, z = (11.5 − 7)/1.8708 = 2.41, and then
P = 2P(X ≥ 12 | p = 1/2) = 2P(Z > 2.41) = (2)(0.0080) = 0.0160 < 0.05.
Decision: Reject H0.
16.7 The hypotheses
H0 : ˜μ2 − ˜μ1 = 8
H1 : ˜μ2 − ˜μ1 < 8.
α = 0.05.
Test statistic: binomial variable X with p = 1/2.
Computations: We have n = 13 and x = 4. Therefore, μ = np = (13)(1/2) = 6.5 and
σ =
p
(13)(1/2)(1/2) = 1.803. Hence, z = (4.5 − 6.5)/1.803 = −1.11, and then
P = P(X ≥ 4 | p = 1/2) = P(Z < −1.11) = 0.1335 > 0.05.
Decision: Do not reject H0.
16.8 The hypotheses
H0 : ˜μ = 20
H1 : ˜μ > 20.
α = 0.05.
Critical region: w≤11 for n = 10.
Computations:
260 Chapter 16 Nonparametric Statistics
di −3 12 5 −5 8 5 −8 15 6 4
Rank 1 9 4 4 7.5 4 7.5 10 6 2
Therefore, w=12.5.
Decision: Do not reject H0.
16.9 The hypotheses
H0 : ˜μ = 12
H1 : ˜μ 6= 12.
α = 0.02.
Critical region: w≤20 for n = 15.
Computations:
di −3 1 −2 −1 6 4 1 2 −1 3 −3 1 2 −1 2
Rank 12 3.5 8.5 3.5 15 14 3.5 8.5 3.5 12 12 3.5 8.5 3.5 8.5
Now, w=43 and w+ = 77, so that w = 43.
Decision: Do not reject H0.
16.10 The hypotheses
H0 : ˜μ1 − ˜μ2 = 0
H1 : ˜μ1 − ˜μ2 < 0.
α = 0.02.
Critiral region: w+ ≤ 1 for n = 5.
Computations:
Pair 1 2 3 4 5
di −5 −2 1 −4 2
Rank 5 2.5 1 4 2.5
Therefore, w+ = 3.5.
Decision: Do not reject H0.
16.11 The hypotheses
H0 : ˜μ1 − ˜μ2 = 4.5
H1 : ˜μ1 − ˜μ2 < 4.5.
α = 0.05.
Critiral region: w+ ≤ 11.
Computations:
Solutions for Exercises in Chapter 16 261
Woman 1 2 3 4 5 6 7 8 9 10
di −1.5 5.4 3.6 6.9 5.5 2.7 2.3 3.4 5.9 0.7
di − d0 −6.0 0.9 −0.9 2.4 1.0 −1.8 −2.2 −1.1 1.4 −3.8
Rank 10 1.5 1.5 8 3 6 7 4 5 9
Therefore, w+ = 17.5.
Decision: Do not reject H0.
16.12 The hypotheses
H0 : ˜μA − ˜μB = 0
H1 : ˜μA − ˜μB > 0.
α = 0.01.
Critiral region: z > 2.575.
Computations:
Day 1 2 3 4 5 6 7 8 9 10
di 2 6 3 5 8 −3 8 1 6 −3
Rank 4 15.5 7.5 13 19.5 7.5 19.5 1.5 15.5 7.5
Day 11 12 13 14 15 16 17 18 19 20
di 4 6 6 2 −4 3 7 1 −2 4
Rank 11 15.5 15.5 4 11 7.5 18 1.5 4 11
Now w = 180, n = 20, μW+ = (20)(21)/4 = 105, and σW+ =
p
(20)(21)(41)/24 =
26.786. Therefore, z = (180 − 105)/26.786 = 2.80
Decision: Reject H0; on average, Pharmacy A fills more prescriptions than Pharmacy
B.
16.13 The hypotheses
H0 : ˜μ1 − ˜μ2 = 8
H1 : ˜μ1 − ˜μ2 < 8.
α = 0.05.
Critiral region: z < −1.645.
Computations:
di 6 9 3 5 8 9 4 10
di − d0 −2 1 −5 −3 0 1 −4 2
Rank 4.5 1.5 10.5 7.5 − 1.5 9 4.5
di 8 2 6 3 1 6 8 11
di − d0 0 −6 −2 −5 −7 −2 0 3
Rank − 12 4.5 10.5 13 4.5 − 7.5
262 Chapter 16 Nonparametric Statistics
Discarding zero differences, we have w+ = 15, n = 13, μW+ = (13)(14)/4 = 45, 5, and
σW+ =
p
(13)(14)(27)/24 = 15.309. Therefore, z = (15 − 45.5)/14.309 = −2.13
Decision: Reject H0; the average increase is less than 8 points.
16.14 The hypotheses
H0 : ˜μA − ˜μB = 0
H1 : ˜μA − ˜μB 6= 0.
α = 0.05.
Critiral region: w ≤ 21 for n = 14.
Computations:
di 0.09 0.08 0.12 0.06 0.13 −0.06 0.12
Rank 7 5.5 10 2.5 12 2.5 10
di 0.11 0.12 −0.04 0.08 0.15 0.07 0.14
Rank 8 10 1 5.5 14 4 13
Hence, w+ = 101.5, w− = 3.5, so w = 3.5.
Decision: Reject H0; the different instruments lead to different results.
16.15 The hypotheses
H0 : ˜μB = ˜μA
H1 : ˜μB < ˜μA.
α = 0.05.
Critiral region: n1 = 3, n2 = 6 so u1 ≤ 2.
Computations:
Original data 1 7 8 9 10 11 12 13 14
Rank 1 2∗ 3∗ 4 5∗ 6 7 8 9
Now w1 = 10 and hence u1 = 10 − (3)(4)/2 = 4
Decision: Do not reject H0; the claim that the tar content of brand B cigarettes is
lower than that of brand A is not statistically supported.
16.16 The hypotheses
H0 : ˜μ1 = ˜μ2
H1 : ˜μ1 < ˜μ2.
α = 0.05.
Critiral region: u1 ≤ 2.
Computations:
Solutions for Exercises in Chapter 16 263
Original data 0.5 0.9 1.4 1.9 2.1 2.8 3.1 4.6 5.3
Rank 1∗ 2 3 4∗ 5 6∗ 7∗ 8 9
Now w1 = 18 and hence u1 = 18 − (4)(5)/2 = 8
Decision: Do not reject H0.
16.17 The hypotheses
H0 : ˜μA = ˜μB
H1 : ˜μA > ˜μB.
α = 0.01.
Critiral region: u2 ≤ 14.
Computations:
Original data 3.8 4.0 4.2 4.3 4.5 4.5 4.6 4.8 4.9
Rank 1∗ 2∗ 3∗ 4∗ 5.5∗ 5.5∗ 7 8∗ 9∗
Original Data 5.0 5.1 5.2 5.3 5.5 5.6 5.8 6.2 6.3
Rank 10 11 12 13 14 15 16 17 18
Now w2 = 50 and hence u2 = 50 − (9)(10)/2 = 5
Decision: Reject H0; calculator A operates longer.
16.18 The hypotheses
H0 : ˜μ1 = ˜μ2
H1 : ˜μ1 6= ˜μ2.
α = 0.01.
Critiral region: u ≤ 27.
Computations:
Original data 8.7 9.3 9.5 9.6 9.8 9.8 9.8 9.9 9.9 10.0
Rank 1∗ 2 3∗ 4 6∗ 6∗ 6∗ 8.5∗ 8.5 10
Original Data 10.1 10.4 10.5 10.7 10.8 10.9 11.0 11.2 11.5 11.8
Rank 11∗ 12 13∗ 14 15∗ 16 17∗ 18∗ 19 20
Here “∗” is for process 2. Now w1 = 111.5 for process 1 and w2 = 98.5 for process 2.
Therefore, u1 = 111.5 − (10)(11)/2 = 56.5 and u2 = 98.5 − (10)(11)/2 = 43.5, so that
u = 43.5.
Decision: Do not reject H0.
16.19 The hypotheses
H0 : ˜μ1 = ˜μ2
H1 : ˜μ1 6= ˜μ2.
264 Chapter 16 Nonparametric Statistics
α = 0.05.
Critiral region: u ≤ 5.
Computations:
Original data 64 67 69 75 78 79 80 82 87 88 91 93
Rank 1 2 3∗ 4 5∗ 6 7∗ 8 9∗ 10 11∗ 12
Now w1 = 35 and w2 = 43. Therefore, u1 = 35−(5)(6)/2 = 20 and u2 = 43−(7)(8)/2 =
15, so that u = 15.
Decision: Do not reject H0.
16.20 The hypotheses
H0 : ˜μ1 = ˜μ2
H1 : ˜μ1 6= ˜μ2.
α = 0.05.
Critiral region: Z < −1.96 or z > 1.96.
Computations:
Observation 12.7 13.2 13.6 13.6 14.1 14.1 14.5 14.8 15.0 15.0 15.4
Rank 1∗ 2 3.5∗ 3.5 5.5∗ 5.5 7 8 9.5∗ 9.5 11.5∗
Observation 15.4 15.6 15.9 15.9 16.3 16.3 16.3 16.3 16.5 16.8 17.2
Rank 11.5 13∗ 14.5∗ 14.5 17.5∗ 17.5∗ 17.5 17.5 20 21∗ 22
Observation 17.4 17.7 17.7 18.1 18.1 18.3 18.6 18.6 18.6 19.1 20.0
Rank 23 24.5 24.5∗ 26.5∗ 26.5 28 30 30∗ 30 32 33
Now w1 = 181.5 and u1 = 181.5 − (12)(13)/2 = 103.5. Then with μU1 = (21)(12)/2 =
126 and σU1 =
p
(21)(12)(34)/12 = 26.721, we find z = (103.5−126)/26.721 = −0.84.
Decision: Do not reject H0.
16.21 The hypotheses
H0 : Operating times for all three calculators are equal.
H1 : Operating times are not all equal.
α = 0.01.
Critiral region: h > χ2
0.01 = 9.210 with v = 2 degrees of freedom.
Computations:
Solutions for Exercises in Chapter 16 265
Ranks for Calculators
A B C
4 8.5 15
12 7 18
1 13 10
2 11 16
6 8.5 14
r1 = 25 5 17
3 r3 = 90
r2 = 56
Now h = 12
(18)(19)
h
252
5 + 562
7 + 902
6
i
− (3)(19) = 10.47.
Decision: Reject H0; the operating times for all three calculators are not equal.
16.22 Kruskal-Wallis test (Chi-square approximation)
h =
12
(32)(33)
210.52
9
+
1892
8
+
128.52
15
− (3)(33) = 20.21.
χ2
0.05 = 5.991 with 2 degrees of freedom. So, we reject H0 and claim that the mean
sorptions are not the same for all three solvents.
16.23 The hypotheses
H0 : Sample is random.
H1 : Sample is not random.
α = 0.1.
Test statistics: V , the total number of runs.
Computations: for the given sequence we obtain n1 = 5, n2 = 10, and v = 7. Therefore,
from Table A.18, the P-value is
P = 2P(V ≤ 7 when H0 is true) = (2)(0.455) = 0.910 > 0.1
Decision: Do not reject H0; the sample is random.
16.24 The hypotheses
H0 : Fluctuations are random.
H1 : Fluctuations are not random.
α = 0.05.
Test statistics: V , the total number of runs.
Computations: for the given sequence we find ˜x = 0.021. Replacing each measurement
266 Chapter 16 Nonparametric Statistics
by the symbol “+” if it falls above 0.021 and by the symbol “−” if it falls below 0.021
and omitting the two measurements that equal 0.021, we obtain the sequence
− − − − − + + + + +
for which n1 = 5, n2 = 5, and v = 2. Therefore, the P-value is
P = 2P(V ≤ 2 when H0 is true) = (2)(0.008) = 0.016 < 0.05
Decision: Reject H0; the fluctuations are not random.
16.25 The hypotheses
H0 : μA = μB
H1 : μA > μB.
α = 0.01.
Test statistics: V , the total number of runs.
Computations: from Exercise 16.17 we can write the sequence
B B B B B B A B B A A B A A A A A A
for which n1 = 9, n2 = 9, and v = 6. Therefore, the P-value is
P = P(V ≤ 6 when H0 is true) = 0.044 > 0.01
Decision: Do not reject H0.
16.26 The hypotheses
H0 : Defectives occur at random.
H1 : Defectives do not occur at random.
α = 0.05.
Critical region: z < −1.96 or z > 1.96.
Computations: n1 = 11, n2 = 17, and v = 13. Therefore,
μV =
(2)(11)(17)
28
+ 1 = 14.357,
σ2
V =
(2)(11)(17)[(2)(11)(17)− 11 − 17]
(282)(27)
= 6.113,
and hence σV = 2.472. Finally,
z = (13 − 14.357)/2.472 = −0.55.
Decision: Do not reject H0.
Solutions for Exercises in Chapter 16 267
16.27 The hypotheses
H0 : Sample is random.
H1 : Sample is not random.
α = 0.05.
Critical region: z < −1.96 or z > 1.96.
Computations: we find ¯x = 2.15. Assigning “+” and “−” signs for observations above
and below the median, respectively, we obtain n1 = 15, n2 = 15, and v = 19. Hence,
μV =
(2)(15)(15)
30
+ 1 = 16,
σ2
V =
(2)(15)(15)[(2)(15)(15)− 15 − 15]
(302)(29)
= 7.241,
which yields σV = 2.691. Therefore,
z = (19 − 16)/2.691 = 1.11.
Decision: Do not reject H0.
16.28 1 − γ = 0.95, 1 − α = 0.85. From Table A.20, n = 30.
16.29 n = 24, 1 − α = 0.90. From Table A.20, 1 − γ = 0.70.
16.30 1 − γ = 0.99, 1 − α = 0.80. From Table A.21, n = 21.
16.31 n = 135, 1 − α = 0.95. From Table A.21, 1 − γ = 0.995.
16.32 (a) Using the computations, we have
Student Test Exam di
L.S.A. 4 4 0
W.P.B. 10 2 8
R.W.K. 7 8 −1
J.R.L. 2 3 −1
J.K.L. 5 6.5 −1.5
D.L.P. 9 6.5 2.5
B.L.P. 3 10 −7
D.W.M. 1 1 0
M.N.M. 8 9 −1
R.H.S. 6 5 −1
rS = 1 −
(6)(125.5)
(10)(100 − 1)
= 0.24.
268 Chapter 16 Nonparametric Statistics
(b) The hypotheses
H0 : ρ = 0
H1 : ρ > 0
α = 0.025.
Critical region: rS > 0.648.
Decision: Do not reject H0.
16.33 (a) Using the following
Ranks Ranks
x y d x y d
1 6 −5 14 12 2
2 1 1 15 2 13
3 16 −13 16 6 10
4 9.5 −5.5 17 13.5 3.5
5 18.5 −13.5 18 13.5 4.5
6 23 −17 19 16 3
7 8 −1 20 23 −3
8 3 5 21 23 −2
9 9.5 −0.5 22 23 −1
10 16 −6 23 18.5 4.5
11 4 7 24 23 1
12 20 −8 25 6 19
13 11 2
we obtain rS = 1 − (6)(1586.5)
(25)(625−1) = 0.39.
(b) The hypotheses
H0 : ρ = 0
H1 : ρ 6= 0
α = 0.05.
Critical region: rS < −0.400 or rs > 0.400.
Decision: Do not reject H0.
16.34 The numbers come up as follows
Ranks Ranks Ranks
x y d x y d x y d
3 7 −4 4 6 −2 7 3 4
6 4.5 1.5 8 2 6 5 4.5 0.5
2 8 −6 1 9 −8 9 1 8
Solutions for Exercises in Chapter 16 269
X
d2 = 238.5, rS = 1 −
(6)(238.5)
(9)(80)
= −0.99.
16.35 (a) We have the following table:
Weight Chest Size di Weight Chest Size di Weight Chest Size di
3 6 −3 1 1 0 8 8 0
9 9 0 4 2 2 7 3 4
2 4 −2 6 7 −1 5 5 0
rS = 1 −
(6)(34)
(9)(80)
= 0.72.
(b) The hypotheses
H0 : ρ = 0
H1 : ρ > 0
α = 0.025.
Critical region: rS > 0.683.
Decision: Reject H0 and claim ρ > 0.
16.36 The hypotheses
H0 : ρ = 0
H1 : ρ 6= 0
α = 0.05.
Critical region: rS < −0.683 or rS > 0.683.
Computations:
Manufacture A B C D E F G H I
Panel rating 6 9 2 8 5 1 7 4 3
Price rank 5 1 9 8 6 7 2 4 3
di 1 8 −7 0 −1 −6 5 0 0
Therefore, rS = 1 − (6)(176)
(9)(80) = −0.47.
Decision: Do not reject H0.
16.37 (a)
P
d2 = 24, rS = 1 − (6)(24)
(8)(63) = 0.71.
(b) The hypotheses
H0 : ρ = 0
H1 : ρ > 0
270 Chapter 16 Nonparametric Statistics
α = 0.05.
Critical region: rS > 0.643.
Computations: rS = 0.71.
Decision: Reject H0, ρ > 0.
16.38 (a)
P
d2 = 1828, rS = 1 − (6)(1828)
(30)(899) = 0.59.
(b) The hypotheses
H0 : ρ = 0
H1 : ρ 6= 0
α = 0.05.
Critical region: rS < −0.364 or rS > 0.364.
Computations: rS = 0.59.
Decision: Reject H0, ρ 6= 0.
16.39 (a) The hypotheses
H0 : μA = μB
H1 : μA 6= μB
Test statistic: binomial variable X with p = 1/2.
Computations: n = 9, omitting the identical pair, so x = 3 and P-value is
P = P(X ≤ 3) = 0.2539.
Decision: Do not reject H0.
(b) w+ = 15.5, n = 9.
Decision: Do not reject H0.
16.40 The hypotheses:
H0 : μ1 = μ2 = μ3 = μ4.
H1 : At least two of the means are not equal.
α = 0.05.
Critical region: h > χ2
0.05 = 7.815 with 3 degrees of freedom.
Computaions:
Ranks for the Laboratories
A B C D
7 18 2 12
15.5 20 3 10.5
13.5 19 4 13.5
8 9 1 15.5
6 10.5 5 17
r1 = 50 r2 = 76.5 r3 = 15 r4 = 68.5
Solutions for Exercises in Chapter 16 271
Now
h =
12
(20)(21)
502 + 76.52 + 152 + 68.52
5
− (3)(21) = 12.83.
Decision: Reject H0.
16.41 The hypotheses:
H0 : μ29 = μ54 = μ84.
H1 : At least two of the means are not equal.
Kruskal-Wallis test (Chi-squared approximation)
h =
12
(12)(13)
62
3
+
382
5
+
342
4
− (3)(13) = 6.37,
with 2 degrees of freedom. χ2
0.05 = 5.991.
Decision: reject H0. Mean nitrogen loss is different for different levels of dietary protein.
Chapter 17
Statistical Quality Control
17.1 Let Y = X1 + X2 + · · · + Xn. The moment generating function of a Poisson random
variable is given by MX(t) = eμ(et−1). By Theorem 7.10,
MY (t) = eμ1(et−1) · eμ2(et−1) · · · eμn(et−1) = e(μ1+μ2+···+μn)(et−1),
which we recognize as the moment generating function of a Poisson random variable
with mean and variance given by
Pn
i=1
μi.
17.2 The charts are shown as follows.
2.385 0 5 10 15 20
2.390
2.395
2.400
2.405
2.410
2.415
2.420
UCL
LCL
Sample
X−bar
00 5 10 15 20
0.003
0.006
0.009
0.012
0.015 UCL
LCL
Sample
R
Although none of the points in R-chart is outside of the limits, there are many values
fall outside control limits in the ¯X -chart.
17.3 There are 10 values, out of 20, fall outside the specification ranges. So, 50% of the
units produced by this process will not confirm the specifications.
17.4 ¯¯X = 2.4037 and ˆσ = ¯R
d2
= 0.006935
2.326 = 0.00298.
17.5 Combining all 35 data values, we have
¯¯
x = 1508.491, ¯R = 11.057,
273
274 Chapter 17 Statistical Quality Control
so for ¯X -chart, LCL = 1508.491 − (0.577)(11.057) = 1502.111, and UCL = 1514.871;
and for R-chart, LCL = (11.057)(0) = 0, and UCL = (11.057)(2.114) = 23.374. Both
charts are given below.
1485 0 10 20 30
1490
1495
1500
1505
1510
1515
1520
1525
LCL
UCL
Sample
X
0 10 20 30
5
10
15
20
25
UCL
LCL = 0
Sample
Range
The process appears to be out of control.
17.6
β = P(Z < 3 − 1.5√5) − P(Z < −3 − 1.5√5)
= P(Z < −0.35) − P(Z < −6.35) ≈ 0.3632.
So,
E(S) = 1/(1 − 0.3632) = 1.57, and σS =
p
β(1 − β)2 = 0.896.
17.7 From Example 17.2, it is known than
LCL = 62.2740, and UCL = 62.3771,
for the ¯X -chart and
LCL = 0, and UCL = 0.0754,
for the S-chart. The charts are given below.
0 10 20 30
62.26
62.28
62.30
62.32
62.34
62.36
62.38
62.40
62.42
LCL
UCL
Sample number
X
0 10 20 30
0.01
0.03
0.05
0.07
0.09
UCL
LCL
Sample number
S
Solutions for Exercises in Chapter 17 275
The process appears to be out of control.
17.8 Based on the data, we obtain ˆp = 0.049, LCL = 0.049 − 3
q
(0.049)(0.951)
50 = −0.043, and
LCL = 0.049 + 3
q
(0.049)(0.951)
50 = 0.1406. Based on the chart shown below, it appears
that the process is in control.
00 5 10 15 20
0.03
0.06
0.09
0.12
0.15
UCL
LCL
Sample
p
17.9 The chart is given below.
00 5 10 15 20 25 30
0.03
0.06
0.09
0.12
0.15
UCL
LCL
Sample
p
Although there are a few points closed to the upper limit, the process appears to be
in control as well.
17.10 We use the Poisson distribution. The estimate of the parameter λ is ˆλ = 2.4. So, the
control limits are LCL = 2.4 − 3√2.4 = −2.25 and UCL = 2.4 + 3√2.4 = 7.048. The
control chart is shown below.
276 Chapter 17 Statistical Quality Control
00 5 10 15 20
1
2
3
4
5
6
7
8
LCL
UCL
Sample
Number of Defect
The process appears in control.
Chapter 18
Bayesian Statistics
18.1 For p = 0.1, b(2; 2, 0.1) =
2
2
(0.1)2 = 0.01.
For p = 0.2, b(2; 2, 0.2) =
2
2
(0.2)2 = 0.04. Denote by
A : number of defectives in our sample is 2;
B1 : proportion of defective is p = 0.1;
B2 : proportion of defective is p = 0.2.
Then
P(B1|A) =
(0.6)(0.01)
(0.6)(0.01) + (0.4)(0.04)
= 0.27,
and then by subtraction P(B2|A) = 1−0.27 = 0.73. Therefore, the posterior distribution
of p after observing A is
p 0.1 0.2
π(p|x = 2) 0.27 0.73
for which we get p∗ = (0.1)(0.27) + (0.2)(0.73) = 0.173.
18.2 (a) For p = 0.05, b(2; 9, 0.05) =
9
2
(0.05)2(0.95)7 = 0.0629.
For p = 0.10, b(2; 9, 0.10) =
9
2
(0.10)2(0.90)7 = 0.1722.
For p = 0.15, b(2; 9, 0.15) =
9
2
(0.15)2(0.85)7 = 0.2597.
Denote the following events:
A : 2 drinks overflow;
B1 : proportion of drinks overflowing is p = 0.05;
B2 : proportion of drinks overflowing is p = 0.10;
B3 : proportion of drinks overflowing is p = 0.15.
Then
P(B1|A) =
(0.3)(0.0629)
(0.3)(0.0629) + (0.5)(0.1722) + (0.2)(0.2597)
= 0.12,
P(B2|A) =
(0.5)(0.1722)
(0.3)(0.0629) + (0.5)(0.1722) + (0.2)(0.2597)
= 0.55,
277
278 Chapter 18 Bayesian Statistics
and P(B3|A) = 1 − 0.12 − 0.55 = 0.33. Hence the posterior distribution is
p 0.05 0.10 0.15
π(p|x = 2) 0.12 0.55 0.33
(b) p∗ = (0.05)(0.12) + (0.10)(0.55) + (0.15)(0.33) = 0.111.
18.3 (a) Let X = the number of drinks that overflow. Then
f(x|p) = b(x; 4, p) =
4
x
px(1 − p)4−x, for x = 0, 1, 2, 3, 4.
Since
f(1, p) = f(1|p)π(p) = 10
4
1
p(1 − p)3 = 40p(1 − p)3, for 0.05 < p < 0.15,
then
g(1) = 40
Z 0.15
0.05
p(1 − p)3 dp = −2(1 − p)4 (4p + 1)|0.15
0.05 = 0.2844,
and
π(p|x = 1) = 40p(1 − p)3/0.2844.
(b) The Bayes estimator
p∗ =
40
0.2844
Z 0.15
0.05
p2(1 − p)3 dp
=
40
(0.2844)(60)
p3 (20 − 45p + 36p2 − 10p3)
0
.
1
5
0.05 = 0.106.
18.4 Denote by
A : 12 condominiums sold are units;
B1 : proportion of two-bedroom condominiums sold 0.60;
B2 : proportion of two-bedroom condominiums sold 0.70.
For p = 0.6, b(12; 15, 0.6) = 0.0634 and for p = 0.7, b(12; 15, 0.7) = 0.1701. The prior
distribution is given by
p 0.6 0.7
π(p) 1/3 2/3
So, P(B1|A) = (1/3)(0.0634)
(1/3)(0.0634)+(2/3)(0.1701) = 0.157 and P(B2|A) = 1 − 0.157 = 0.843.
Therefore, the posterior distribution is
Solutions for Exercises in Chapter 18 279
p 0.6 0.7
π(p|x = 12) 0.157 0.843
(b) The Bayes estimator is p∗ = (0.6)(0.157) + (0.7)(0.843) = 0.614.
18.5 n = 10, ¯x = 9, σ = 0.8, μ0 = 8, σ0 = 0.2, and z0.025 = 1.96. So,
μ1 =
(10)(9)(0.04) + (8)(0.64)
(10)(0.04) + 0.64
= 8.3846, σ1 =
s
(0.04)(0.64)
(10)(0.04) + 0.64
= 0.1569.
To calculate Bayes interval, we use 8.3846 ± (1.96)(0.1569) = 8.3846 ± 0.3075 which
yields (8.0771, 8.6921). Hence, the probability that the population mean is between
8.0771 and 8.6921 is 95%.
18.6 n = 30, ¯x = 24.90, s = 2.10, μ0 = 30 and σ0 = 1.75.
(a) μ∗ = n¯x 2
0+μ0 2
n 2
0+ 2 = 2419.988
96.285 = 25.1336.
(b) σ∗ =
q
2
0 2
n 2
0+ 2 =
q
13.5056
96.285 = 0.3745, and z0.025 = 1.96. Hence, the 95% Bayes
interval is calculated by 25.13±(1.96)(0.3745) which yields $23.40 < μ < $25.86.
(c) P(24 < μ < 26) = P
24−25.13
0.3745 < Z < 26−25.13
0.3745
= P(−3.02 < Z < 2.32) =
0.9898 − 0.0013 = 0.9885.
18.7 (a) P(71.8 < μ < 73.4) = P
71.8−72 √5.76
< Z < 73.4−72 √5.76
= P(−0.08 < Z < 0.58) =
0.2509.
(b) n = 100, ¯x = 70, s2 = 64, μ0 = 72 and σ2
0 = 5.76. Hence,
μ1 =
(100)(70)(5.76) + (72)(64)
(100)(5.76) + 64
= 70.2,
σ1 =
s
(5.76)(64)
(100)(5.76) + 64
= 0.759.
Hence, the 95% Bayes interval can be calculated as 70.2 ± (1.96)(0.759) which
yields 68.71 < μ < 71.69.
(c) P(71.8 < μ < 73.4) = P
71.8−70.2
0.759 < Z < 73.4−70.2
0.759
= P(2.11 < Z < 4.22) =
0.0174.
18.8 Multiplying the likelihood function
f(x1, x2, . . . , xn|μ) =
1
(2π)25/210025 exp
"
−
1
2
X25
i=1
xi − μ
100
2
#
by the prior π(μ) = 1
60 for 770 < μ < 830, we obtain
f(x1, x2, . . . , xn, μ) =
1
(60)(2π)25/210025 exp
"
−
1
2
X25
i=1
xi − μ
100
2
#
= Ke−1
2 (μ−780
20 )2
,
280 Chapter 18 Bayesian Statistics
where K is a function of the sample values. Since the marginal distribution
g(x1, x2, . . . , xn) = √2π(20)K
1
√2π20
Z 830
770
e−1
2 ( μ−780
100 )2
dμ
= √2π(13.706)K.
Hence, the posterior distribution
π(μ|x1, x2, . . . , xn) =
f(x1, x2, . . . , xn)
g(x1, x2, . . . , xn)
=
1
√2π(13.706)
e−1
2 ( μ−780
20 )2
,
for 770 < μ < 830.
18.9 Multiplying the likelihood function and the prior distribution together, we get the joint
density function of θ as
f(t1, t2, . . . , tn, θ) = 2θn exp
"
−θ
Xn
i=1
ti + 2
!#
, for θ > 0.
Then the marginal distribution of (T1, T2, . . . , Tn) is
g(t1, t2, . . . , tn) = 2
Z
∞
0
θn exp
"
−θ
Xn
i=1
ti + 2
!#
dθ
=
2(n + 1)
Pn
i=1
ti + 2
n+1
Z
∞
0
θn exp
−θ
Pn
i=1
ti + 2
(n + 1)
Pn
i=1
ti + 2
−(n+1) dθ
=
2(n + 1)
Pn
i=1
ti + 2
n+1 ,
since the integrand in the last term constitutes a gamma density function with parameters
α = n + 1 and β = 1/
Pn
i=1
ti + 2
. Hence, the posterior distribution of θ
is
π(θ|t1, . . . , tn) =
f(t1, . . . , tn, θ)
g(t1, . . . , tn)
=
Pn
i=1
ti + 2
n+1
(n + 1)
θn exp
"
−θ
Xn
i=1
ti + 2
!#
,
for θ > 0, which is a gamma distribution with parameters α = n + 1 and β =
1/
Pn
i=1
ti + 2
.
Solutions for Exercises in Chapter 18 281
18.10 Assume that p(xi|λ) = e− xi
xi! , xi = 0, 1, . . . , for i = 1, 2, . . . , n and π(λ) = 1
24 λ2e− /2,
for λ > 0. The posterior distribution of λ is calculated as
π(λ|x1, . . . , xn) =
e−(n+1/2) λ
n P
i
=
1
xi+2
R
∞
0 e−(n+1/2) λ
n P
i
=
1
xi+2
dλ
=
(n + 1/2)n¯x+3
(n¯x + 3)
λ
n P
i
=
1
xi+2
e−(n+1/2) ,
which is a gamma distribution with parameters α = n¯x + 3 and β = (n + 1/2)−1,
with mean n¯x+3
n+1/2 . Hence, plug the data in we obtain the Bayes estimator of λ, under
squared-error loss, is λ∗ = 57+3
10+1/2 = 5.7143.
18.11 The likelihood function of p is
x−1
4
p5(1−p)x−5 and the prior distribution is π(p) = 1.
Hence the posterior distribution of p is
π(p|x) =
p5(1 − p)x−5
R 1
0 p5(1 − p)x−5 dp
=
(x + 2)
(6)(x − 4)
p5(1 − p)x−5,
which is a Beta distribution with parameters α = 6 and β = x − 4. Hence the Bayes
estimator, under the squared-error loss, is p∗ = 6
x+2 .
Posts
