Answers to Selected Problems

Chapter 2
1. (a) Y e s , (b) No, P r ( E , ) < O, (c) No, r P r ( E ¡ ) < 1, (d) No, r P r ( E ¡ ) > 1.
2. (a) Pr(Á) = . 4 5 , Pr(C) = . 7, Pr(A U C) = . 6 5 , Pr(B U C) = .65,
Pr(A U B U C) = 1. (b) Pr(A U B) = . 90 = . 55 i- . 35, since A and B
are m. e. events. (c) Pr(A U C) = . 65 / . 55 + . 30, since A and C are
not m. e. events.
3. No, not necessarily. This result would be true if the probability model
is uniform.
4. (a) Pr(Á) = 1/2, Pr(C) = 2/3, P r ( A U C ) = 2 / 3 , P r ( B U C ) = 2 / 3 ,
Pr(A U B UC) = 1.
5. (a) C o n s i d e r t h e event "one boy." In S t the event "one boy" is composed
of {BGG, GBG, GGB}. Henee, under S|, Pr(one boy) = 3/8 whüe
usíng S 2 , Pr(one boy) = 1/4. These resulta are not consistent.
(b) Observe a large number of families witb three children and see
which assignment of probability aeems more reasonable.
(c) SLnce from S¡ to S2 is a raany-to-one correspondence, assignLng
probabilities in S, uniquely determines the probabüities in S 2 . In
contrast, S 2 to S, has a one-to-many correspondence. Henee for a
given assignment in S 2 , many different probability assignments in
S[ are poasíble.
6. (a) (i) red
1 2 3 4 5 6
green
1 (1, 1) (1, 2) (1, 3) (1, 4) (1, 5) (1, 6)
2 (2, 1) (2, 2) . . . (2, 6)
3
4

5
6 (6, 1) . . . . (6, 6)

285
286 Answers Answers 287

(¡i) S = {(g, r) I g = 1, 2, . . . , 6 and r = 1, 2 6}. (b) P r f n o boy or more than one boy} = 5/8, Pr{no boy} = 1/8,
(b) A = { ( 1 , 1 ) ( 2 , 2) (3, 3) (4, 4 ) ( 5 , 5) (6 , 6)}or P r j a t least two boys} = 1/2
(c) (iv) and (vi)
A = {(g, r) I s = r for g = 1, 2, 3, . . . , 6 and (d) Prjexactly one boy} = 3/8, Pr{more boys than girla} = 1/2,
r = 1, 2, . . ., 6}, B = {(g, r) | g > r + 2}, Pr{orúy the eldest child is a boy} = 1/8
(e) (iv) and ( v i )
C = {(g, r) I g + r = 8}.
24.
(c) (i) The number on the red die is rwice that on green.
(¡i) The sum ¡3 greater thají 5. 26. (a) Yes, yes.
(¡u) The two outcomes are difieren!. (b) Q = (Q n C , ) U (Q O C 2 ) U • • • U (Q n C 3 5 ) . We would need
(d) Pr(A) = 1/6, Pr(B) = 5/18, Pr(C) = 5/36, Pr(r = 2g) = 1/12, Pr(Q n C¡) for i = 1, 2 35.
Pr(r + g > 5) = 13/18, Pr(r * g) = 5/6. 27. (a) S = {OGP, OGP, OGP, OGP, ÓGP, ÓGP, ÓGP, ÓGP} or
7. 18/36 S = {E 1 , E 2 , E 3 , E 4 , E 5 , E 6 > E 7 , Ej
8. (a) HLnf. B = (B CÍA) U (B fl A) = ( B O A ) U A (d) Pr(E¡) = 1/8 for i = 1, 2, . . . , 8
13. (a) S = {(x, y) | x = O, 1 5 and y = O, 1, 2, 3} Chapter 3
(d) 3/8, 5/12, 1/6, 1/8, 11/24, 1/24, 1/2, (e) 11/24, 1/2 1. (a) 3024, (b) 6!, (c) n(n - 1), (d) k
14. (a) 2/3, 1/3, 1/2, 1/2, 1/6, 5/6, 2/3, 1/3 2. (a) 120, (b) 1, (c) m(m - l)/2, (d) t, (e) 500(499)/2, (f) k
17. . 13, O, .17, .55, .45, .15 5. (a) 9, (b) 11, (c) 10
18. (b) 1/2, 1/2, 1/5, 4/5, 1/2, 1/5. (c) No change. 6. The number of subsets of size r is the same as the number of subsets
19. (a) S = {R,R 2 , R , W 2 , R,P 2 , W , R 2 , W , W 2 , W , P 2 , P,R 2 , P,W 2 , of size (n - r ) .
(e) Each color has equal representation. 7. (a) 22, 20, (b) 10/11

(f) 5/9, 2/9, 8/9, 1/3, 4/9, 5/9, 8/9, 2/9, 5/9, 5/9 8. (b) 1/4, 1/2, 1/2

20. N Ñ 9. (b) 1/4, 1/2, 1/2

10. (a) 8!, (b) 1/28
D .002 .008 .01
11. 44
b .398 .592 .99
12. 4095
.4 .6 13. 8/17, 4/17, 1/17
21. (a) Le! the event (A, D, B, C) represent A selecting A's turtle, 14. 60 are needed.
B selecting D's turtle, C selecting B's turtle and D selecting C's
turtle. Thus S consista of the 4! = 24 orderings of the letters 15. 190
A, B, C, D. 16. 256, 1/2
(b) 1/24, 3/8, 1/4, 1/12 17. 14!/5!3!3!3!
22. (c) 3/7, (di 3/7 18. 84, 8P4
23. (a) 3/8, 7/8, 1/2, 1/2, 1/2, 1/2
19. 6/15, l/f1,5)
288 Answers Answers 289

20. 10!/4!2!, 1/30, 1/5 3! 10!

*3
2!3!3!4!
21. 1/35, 1/144
43. (a) A typical point in S: (W, M), where each of W and M can be fílled
22. 3071
in f o u r ways. Henee, S contains 16 points. The uniform Tnodel can
23. 1/55, 1/66, 6/55 be applied.
2 4 . 1/126, 1/21, 4/9 (b) Pr(E) = 1/4 and Pr(F) = 1/4. E D F is the selection of any m a r r i e d
couple other than Mr. and Mrs. A.
25. 60, 1/5
44. 3/14, 1/4, 1/6
26. 1/35, 1/35, 1/140
45. 1/4
27. 1/6
/14>//25j m j / ( 2 5 \j Chapter 4
28.
1. (a) Pr(A) = 7/10, P r ( E , | A) = 2/7, P r ( E 2 I A) = O, P r ( E 3 1 A) = 3/14,
29. 77/300, which is not the aame as the answer to 28 (c).
Pr(E 4 I A) = 1/2, Pr(E 5 I A) = O .
30. 2/15, 1/3, 2/3
(b) Since O < Pr(E ( | A) < 1 for all i and £ j Pr(E¡ I A) = 1, this ¡s a
31. 4/25, 9/25, 16/25 proper probability model.
32. 12/49, 1/4, 1/25 2. (a) P r { ( l , 3) | sum = 4} = P r { ( 3 , 1) | sum = 4} = P r { ( 2 , 2) | aum = 4}
= 1/3 and all other points have zero probability.
33.
(b) 2/3, (c) 2/3
34. 1/100, 1/10 3. Pr(E¡) = 1/k for i = 1, 2 k. Let A = EL U E m+1 U • • • U E^+^i
henee Pr(A) = (r + l)/k. Then Pr(E¡ I A) = l/(r + 1 ) if E¡ C A and
35. 30/49
Pr(E¡ | A) = O if E¡ n A = 0, which gives a uniform probability model
36. lA/N with r •*• 1 eventa.
37. (a) A typical point in S is the sLx-tuple (x, , x 2 x¿) where x¡ is 4. (a) 1/2, (b) 1/6
R or B. In all S will have 64 points. 5. (a) 1/2, (b) 1/2
(b) The uniform model is not appropriate since there are not equal
numbers of red and blue balls. 6. (a) A Á
38. Typical point Ln S: (Xj^, x 2 , • • • , x^.) where x¡ can be 1, 2, . . . , 6. B .1 .3
S containa 6^ points. Here the uniform model is appropriate.
.1 .5
39. d(d - l)/k(k - 1)
(b) (i) 1/4, (¡i) 1/2, (¡U) 5/6, (¡V) 9/10
40. There are 5 2 ! / 4 ! 4 8 ! points in S.
7. (c) Pr(three heada) = 1/8, Pr(three heada I at leaat two heads) = 1/4.
/46\//52i /5 \2 i / 5 2 - r \ / / 5 2 \1 I (b) ( ) /í }, (c) ( )/ ( ),
/' ^S'/ ^ 4 ' \ / / V 4/ ' 8. Pr(A I B) = 0; in the conditional sample space, A is an irnpossible event.
9. Pr(A i B) = 1; Ln the conditional sample space, A is a certain event. Lf
we know B has occurred, A must have occurred.
•
13! 10. (b) 1/24, (c) 1/4, 1/2, 1/8, 5/8, 1/4, 1/2, 1/2, 3/4
41.
4!3!2!2!
290 Answers Answers . 291

11. Pr(A n B O C) = Pr((A n B) n C] = Pr(A n B)Pr(C I A n B) = 28. 53/70, 17/70

Pr(A)Pr(B 1 A ) P r ( C I A n B) . This is proved using successive applica-
30. (b) (i) 119/120, (u) 5/12
tions of the multiplícatlon law.
31. 9/32
12. (a) b/(w + b - 1), (b) w/(w + b), (c) b/(b + w - 1)
32. 1/8
13. (a) b/(w + b), (b) w/(w + b) , (c) b/(w + b)
33. 9/59
14. Let R¡ be an adverse reaction to first shot and R 2 be an adverse reac-
tion to second shot. Then P r ( R , ) = .4, Pr(R, ) = .6, Pr(R 2 | R , ) = O, 34. 3/32
Pr(R 2 | R , ) = .6.
35. Pr(lst | S) = 5/11, Pr(2nd I S) = 3/11
(a) Pr(R 2 D R , ) = Pr(R 2 I R , ) P r ( R , ) = .36
36. (a) 21/23, (b) .87
(b) Pr(R, U R 2 ) = .4 + .36 = .76 sLnce R, and R 2 are m . e .
37. (a) 19/217, (b) error = (+ n D) or (-DD), (c) Pr(error) = .0995
15. E| (3, 0, 0)(0, 3, 0)(0, 0, 3) (1, 2, 0) (2, 1, 0)(0, 2, 1) 38. (a) 95/105, (c) .075
1 1 3
Pr(Et) 27 27 27 27 27 27 Chapter 5
(1, o, 2)(2, 0, D(0, 1, 2) (1, 1, 1) 2. (a) .0001, (b) .9999
3 _3_ 3 6 3. (a), (b), (c) are índependent.
Pr(E,) 27 27 27 27
4. Pr(G) « 1/4, Pr(G I H) = 1/4; G and H are Índependent.
3 21 1
Pr(all three in one box I at least two ¡n same box) = — -i- — = — 6. 9
7. 1/14, 2/7
16. (a) Pr(LL) = .4, Pr(LR) = .l, Pr(RL) = . 3 , Pr(RJ3) = .2
8. 1/6, 1/4; events A and B not Índependent.
(b) Pr(Right on 2) = Pr(LR) + Pr(RR) = .1 + .2 = .3
9. True, only ¡f A and B are m . e .
(c) Pr(R on Ist I R on 2nd) = , 2 / . 3 = 2/3
10. . 2 7 , no, 27/40
11. ( . 8 ) 3 ( . 4 ) 3
17.
12. (a) ( a + b ) / n , (a + c)/n, a/(a + c), a/(a + b)
(b) If ad - bo = O, tben b = ad/c. Consider Pr(M) = (a+ b)/(a + b + c + d).
18. 1/32, 1/16, 1/8, 1/4, 1 Let b = ad/c and show that Pr(M) = a/(a + c), which ts also
Pr(M I E); henee M and E are índependent.
(c) Apply Theorem 5.1.
13. (a) ( . 8 ) 2 ( . 5 ) 2 , (b) ( . 8 ) 2 ( . 5 ) 2 + 4 ( . 8 ) ( . 2 ) ( . 5 ) í + ( . 2 ) 2 ( . 5 ) 2 ,
20. (b) 3/11, 1/3, 3/7
(c) ( . 8 ) 2 ( . 5 ) 2 + 4 ( . 8 ) ( . 2 ) ( . 5 ) 2 + ( . 2 ) 2 ( . 5 ) 2
21. . 7 3 4 7
14. (a) ( . 5 ) 4 , (b) 6 ( . 5 ) 4 , (c) 6 ( . 5 ) 4 .
22. 1/15, 1/2, 16/17
15. Pr(E n F I F) = Pr(E); that is, proportion of outcomes in E O F relative
23. 1/2, 44/91
to F is the same as the proportion of outcomes In E relative to S. This
24. .15, .15 is a result of the índependence of E and F.
25. (b) 5/8, (c) 3/8, (d) 7/10 16. (a) ( . I ) 6 , ( . 9 ) 6 , 6 ( . 1 ) ( . 9 ) 5 , l - ( . 9 ) 6 , 1 - (. 9) 6 - 6(. 1) (. 9) 5
(b) índependence
292 Answers Answers 293

n n n-1 n / n \ k , n-k Chapter 6

17. p , (1 - p) , n p ( l - p) , 1 - (1 - p) , (k)p (1 - p)
1. (a) E¡ H TH TTH TTT
18. (a) p2 + (1 - p)p E , (b) p,p 2 -t- (1 - p , ) p 2 p 3
19. (a) Pr(R 2 | R , ) = 3/8, Pr(R 2 ) = Pr(R, O R 2 ) + P r ( W t fl R 2 )
= 3/8 • 3/8 + 5/8 • 3/8 = 3/8 ;
Pr<E> | (Í> d) (¡)\¡) ar
x 1 2 3 3
therefore, R, and R 2 are independent.
Assumption: Independence
(b) Pr(R 2 | R , ) = 2/7, P r ( R 2 ) = 3/8; therefore R[ and R 2 are not
Lndependent. (b) x 1 2 3
20. 1 - (1 - 6) k 3/8 15/64 25/64
f(x)
23. (a) Pr(A) = 11/16, P r ( B ) = 7 / 1 6 , Pr(C) = 9/16
F(x) 3/8 39/64 1
(b) Pr(A O B) = 4/16 * [11/16][7/16]; henee, A and B are not Lndepend-
ent. A, B and C cannot be mutually Lndependent.
2. (b) m 0 1 2 3
(c) P r ( A | B) = 4/7 * Pr(A)
P r ( M = m) 1/3 1/2 0 1/6
(d) 1/36, 2/36, 5/36, 10/36, 1/36, 5/36, 2/36, 10/36
(c) Pr(M = n) = 1/n ! , Pr(M = n - 1) = 0
25. P r ( I ) = 1/5, P r ( I I ) = 4 / 5 , P r ( R | I ) = 2 / 5 , Pr(R I II) = 1/5,
It is not possible to have (n - 1) matches, since (n - 1) matches
Pr(I | R) = 1/3, Pr(U | R) = 2/3 impliea n matches.

n 3. (a) 1/12, (b) x - 1 1 3

n / S W(2/5)
(1/5) 9 / R \. (a) Pr(I I nR) =

(2/5) n (l/5) +(4/5)(l/5) n 2n w - 1 0 1

(4/5)(l/5) n 4 f(X) 1/4 1/4 1/2

Pr(II | nR) =
4. (b) x 2 3 4 5 6 7 8
(b) n = 2
J_ _2_ 3 4 3 2 1
(c) As n gets large, Pr(I | nR) approachea 1 and Pr(II | nR) approaches f(X)
16 16 16 16 16 16 16
zero; that is, íf all n balls are red, it is nearly certain that they
are being drav/n from box I.
5. (b) x 2 3 4 5 6 7 8
27. (a) 1 4 10 20 25 24 16
f(x)
100 100 100 100 100 100 100
(b) R e l ( I ) > Rel(II) is equivalent to (1 - Pg)(l - P T ) > 0. Thus system I
¡s always more reliable than II unless pg = 1 or p^, = 1 or both Pg 1.5 3
6. (a) x 5
and pT are equal to 1.
29. I: Aa x aa — all Aa f(X) 1/2 1/3 1/6
II: Aa X Aa — 1/1 AA, 1/2 Aa, 1/4 aa. Given an equal nutnber of each
type of mating: (b) y 3 6 10
(a) 1(1/2) + 1/2(1/2) - 3/4, (b) 2/3
f(y) 1/2 1/3 1/6
30. (a) 2/3, (b) 1/2
294 Ansí Answers 295

7. (b) X 4 5 6 7 8 9 (b) The p . f . of X remains the same since each ordering is equally
likely.
f(x) .1 .3 .3 0 .2 .1
m
14. (a) s 1 2 3 4 5 6
(c) •7, 1, .3, .6
8. (a) 9/80, (ci 9/16 2 .1/36 0 0 0 0 0
9. (a) x 0 1 2 3 3 0 2/36 0 0 0 0
4 0 1/36 2/36 0 0 0
f(x) 1/8 3/8 3/8 1/8
5 0 0 2/36 2/36 0 0
(b) y - 3 - 1 1 3
6 0 0 1/36 2/36 2/36 0

g(y) 1/8 3/8 3/8 1/8 7 0 0 0 2/36 2/36 2/36

8 0 0 0 1/36 2/36 2/36
(c) X
0 1 2 3
y 9 0 0 0 0 2/36 2/36
-3 1 / 8 0 0 0 10 0 0 0 0 1/36 2/36
-i 0 3/8 0 0 11 0 0 0 0 0 2/36
i 0 0 3 / 8 0 12 0 0 0 0 0 1/36
3 0 0 0 1 / 8
(b) m 1 2 3 4 5 6
(d) X and Y are not independen!.
g(») _L JL JL . JL _§_ n
10. (a) S = {(x, , x 2 , X 3 , x 4 , x 5 ) x¡ = D or N}. There will be 2 5 = 32 36 36 36 36 36 36
poüits in S.
s 2 3 4 5 6 7 8 9 10 11 12
(b) x 1 2 3 4 5
1 2 3 4 5 6 5 4 3 2 1
h(s)
f(x) 1/16 4/16 6/16 4/16 1/16 36 36 36 36 36 36 36 36 36 36 36
11. (c) Using part (a) : (c) No, since f ( m , s) * g(m)h(s) for all m and s.
x 0 1 2 3 4 5 15. Sample space:
Boxes
f(x) (i/ 5(i; 10(i; 10(i/ 5(i)5 (i; 1 2 3 Probability 1
Boxes
2 3 Probability
Using part (b): 4 0 0 1/81 1 0 3 4/81
X 0 1 2 3 4 5 0 4 0 1/81 3 0 1 4/81
0 0 4 1/81 0 2 2 6/81
g(x) (¡í »(iX 1 i4
3/
fZ \íl \
\S 0T / \ N' OÍ /
/ 2 \ V l i2
V
X 'Oí /NVO
'í/
/ 2 , V l \2 f
5V3/V3/ V3/ 3 1 0 4/81 2 0 2 6/81
1 3 0 4/81 2 2 0 6/81
13. (a) X 0 1 2 3
0 1 3 4/81 1 1 2 12/81
f(X) 2/6 3/6 0 1/6
0 3 1 4/81 2 1 1 12/81
1 2 1 12/81
296 Answers 297
Answers

X
(a) 0 1 2 3 4 19. (b) x [
V
0 0 24/81 12/81 0 0 l \)
f(x)
1 14/81 8/81 12/81 8/81 0

2 2/81 0 0 0 1/81

x 0 1
32
2
24 8
3 4
1
:-/
u/
f(x) g 81
81 81 81 X
20.
y
0 1 2 fz(y)
y 0 1 2
36 42 3 0 (•3)(1) (.4)(.5) (.3)(.25) .575
«w 81 81 81 1 (.3)(0) (.4) (.5) (.3) (.5) .35
(c) No. 2 (.3,(0) (.4)(0) (.3)(.25) .075
16. X
.3 .4 .3
f(x) 3/14 8/14 3/14 X and Y are not independen!.

F(x) 3/14 11/14 14/14 21. (a) S = {(x, y) |x = 3 , 4 , 5; y = l, 2, . . .

X
17. (a) E¡ M FM FFM FFF y 3 4 5
13
Pr(E¡) (1 - p)p (1 - p) 2 p (1 - p)3 1
(¡)d) (í>© (U© 36

1 2 3 2
(¡)d) ®& m 3?
f(g) p (1 - p)p (1 - p) 2 p (1 - p)3
3 0
(IXfJ (¡)& J7_
36

(¡)(Di
(c) y O 1 2 3
4
0 0
h(y) p3 3(1 - p)p 2 3(1 - p) 2 p (1 - p) 3 sé'
1 !_
(d) g =y I O 1 2 3 •3
3 3

1/2 1/4 1/8 1/8 X and Y are not índependent.

f(g)
22. y 1 0 1 3 2 4
1/8 3/6 3/8 1/8
f(y) 1/5 1/5 1/5 1/5 1/5

X 0 1 2 3 4
298 Answers Answers 299

23. (a) y 0 1 (c) E ( Z ) = dE(Y) + c, Var(Z) = d z Var(Y)

(d) E(C) = 5{E(F) - 32J/9, Var(C) = (25/81) Var(F)
f(y) 1/5 2/5 2/5
E(F) = (9/5) E(C) + 32, Var(F) = (81/25) Var(C)
O 1 4. (a) WR: 0 1

f(z) f(x) .216 .432 .288 .064

(c) W = Z 2 ( m o d 5)
WOR: 0 1 2 3

f(x) .167 .500 .300 .033

f(w)
(b) Expected number: WR 21.6 43.2 28.8 6.4
24. p,(x) = x/11, x = 2, 4, 5 p 2 (y) = y/6, y = 1, 2, 3; X and Y are WOR 16.7 50.0 30.0 3.3
independent.
(c) Observed frequencies more closely resemble samplLng WOR.
25. y O 1
5. Z (x. - c)f(x.) = S x.f(x.) - c T, f(x.) = E(X) - c ; henee, tf
-1 1/16 4/16 0 0 0
E (x¡ - c ) f ( X j ) = O, then c = E(X).
O O O 6/16 O O
6. E((X - a ) 2 ] = E { ( X - E(X) + E(X) - a) 2 }
1 0 0 0 4/16 1/16
= E { ( X - E(X)) 2 + 2[X - E(X))[E(X) - a] + [ E ( X ) ' - a] 2 }
X and Y are not independent.
= E{[X - E(X)] 2 } + E{[E(X) - a]2} + 2[E(X) - a)E[X - E(X)]
Chapter 7 = Var(X) + [E(X) - a] 2 + 0.
1. (a) E(X) = O , Var(X) = 1.6 Thus, a equal to E(X) minimizes E { (X - a) 2 }.
(b) y -7 -4 7. E(Y) = a + bE(X) + cEiX 2 ). RecaU that E(X 2 ) = Var(X) + [ E ( X ) ] Z ; thus
E (Y) = a + bE(X) + c Var(X) + c[E(X)] 2 .
f(y) .2 .1 .3 .3 .1
k
E(Y) = -1, Var(Y) = 14.4 10. (a) E(Y) = ^ I x . - E(X)| f(x.)
(c) 3E(X) - ! = -!= E (Y), 9 Var(X) = 9 ( 1 . 6 ) = 14.4 = Var(Y)
2. E(X) = .55, E(Y) = 2 . 1 , Var(X) = 1.3475, Var(Y) = 5.39,
E(Z) = {x. - E(X)} f ( x . )
E(Y) = 2E(X) + 1, Var(Y) = 4 Var(X) í=l
3. (a) ^ E ( Z ) = E(Y) + c = ^ + c Since "absolute valué" and "squarlng" always give positive quanti-
Y = Z - c < ties and f(x¡) >_ O, E(Y) and E(Z) will be sums of positive quantities.

Var(Z) = Var(Y) = ir (b) If Pr(X = E(X)¡ = 1, then E(Y) = O, E(Z) = 0.

(c) E (Y) and E(Z) increase as X deviates from E(X).
(b) (d) Variance of X

Pr(V = V) I ,2 .2 .3 .2 ,1
E(V) = - . 2 , E(X) = 7 6 . 8 , Var(V) = 1.56, Var(X) = 1.56
300 Answers Answers 301

U. 0 1 2 E(S) Var(S)
(c) Var(Z) =
f, (s) .32 .56 12 .8 .4 12

fj (s) (.6)2 2(.4)(.6) (.4)2 .8 .48 1

2 0 . (a) f ( x ) = — for x = 1, 2, . . . , n
n
f3(s) .3588 .4824 .1588 .8 .4776
f, (s) .4 .4 2 .8 .56 F(x) = - for x = 1, 2, . . . , n

fs(s) .398 .404 198 .8 .556

(a) x 1 2 3
(c) fi) Roliing a balanced die with n = 6.
f(X) P P(l - P) (1 - P)2 (i¡) Drawing a random number from 01, 02, 99 with n = 99.
(d) a = 1, b = n - 1
(b) E(X) = p 2 - 3p + 3, V a r ( X ) = p(5 - lOp + 6p 2 - p 3 )
21. New Method: D D
15. (a) 10025, (b) 100, (c) 10, (d) -100, (e) 25, (f) 5
E(profit) = . 5875 Probability .05 .95
16. (a) E ( X ) = p , Var(X) = p(l - p) Profit -.60 .65
(b) Var(X) = p(l - p) has máximum of 1/4 when p = 1/2. Oíd Method:
(c) E ( X 2 ) = E(X) = k, then Var(X) = k - k 2 = k(l - k ) . E(profit) = . 7 5 Probability .08 .92
Var(X) < 1/4 for O < k < 1. Profit -.40 .85

(d) Indlcator random variable. Continué with oíd method.

17. (a) £(X) = 3, Var(X) = 2 22. Assume acb is correct; then

(b) Y abe acb bac bca cab cba

1 3 0 1 1 0
f(y) 1/25 3/25 5/25 7/25 9/25 E(X) = 1
E(Y) = 95/25 = 3.8, Var(Y) = 1. 36 23 • S = {D, GD, GGD, GGG} with probabilities 1/4, 3/16, 9/64, 27/64.
(c) SLnce X = 3 for GGD and G G G , E(X) = 37/16.
25. (a) $ 2 . 5 0 , (b) $3.33
f(y) O 1/10 2/10 3/10 4/10
£(Y) = 4, Var(Y) = 1.0 26. (a) Jack's expected net gain = — ($2) + — ($4) - $1 = -33¿.
J6 ¿b
18. 5287/1024 = 5.16 (b) He should pay 67 cents if the game is to be f a i r .

19. (a) z a a+1 ... a +b 27. (b) E < Y ! ) = 2, E ( Y 2 ) = 1.8

1 28. E ( G , ) = -2/3, V a r ( G , ) = 8/3

í-(z)
b + 1 b + 1 b + 1 E ( G 2 ) = -2/3, V a r ( G 2 ) = 10/3
1 29. X = label on chosen ball. Then
or f ( z ) = for z = a, a + 1, . . . , a
2x
for x = 1, 2,
(a (a + b)} = a + b/2 n(n + 1)
2n + 1 (n + 2)(n - 1)
E(X) = Var(X) =
18
 Answers Answers 303
302

2 _ (b + d)(a + c) 2 (c + d) (a + b)
33. í(t) = for t = 2, 3, .,7 14. (a)
CTX~ 2
36 n ^ n2
13 - t for t = 8, 9, . . . , 12 ad - be
36
E(T) = Med(T) = Mode(T) = 7
15. (a) p xy = O ; X and Y are independent since p = O implies independence
Chapter 8 for indicator raudom variables.
™, O ; X and Y are not independent.
2. (a) E(Z) = E(W) = 3/2, Var(Z) = Var(W) = 3/4 AY

(b) Cov(Z, W) = -3/4; Z and W are not independent.

20. Var(U) = o2(kz + 1) = \s.laz since k is much greater than 1. Var(U) is
( c ) , (d) E(Z + W ) = 3 , Var(Z + W) = O approximately the same as Var(Y); the X component is neglígible.
3. (b) E ( X ) = 2 0 / 3 , E(Y) = 15/7, E(XY) = 58/3 1

21. (a) Var(S) = na2 [1 + (n - l)p], (b) p = -

(c) X and Y are not independent. n - 1

4. (b) E(X + Y ) = 4 , Var(X + Y ) = l , E ( X Y ) = 4 , E(X/Y) = 1.2 22. (b) p ( V , W) = 1/2

(c) E(X) = 2, Var(X) = .4, E (Y) = 2, Var(Y) = .6 23. p(X, Z) = p ( X , X + Y) = "/2/2

(d) E(X + Y) = E(X) + E (Y), E(XY) = E(X) E(Y), E(X/Y) * E(X)/E(Y) 24. (a) E ( X ) = 2 , E(Y) = 6, E(Z) = 2

(e) X and Y are not independent since f(x, y) * f, (x)f 2 (y) for all (x, y ) . (b) Var(X) = 8/5, Var(Y) = 4/5, V a r ( Z ) = 8/5
Var(X + Y) = Var(X) + Var(Y) since Cov(X, Y) = 0. X and Y are independent and so are Y and Z .
5. (a) E(X) = E(Y) = 2/3, Var(X) = Var(Y) = 5/9, Cov(X, Y) =-5/18, (c) E(X + Y + Z) = 10, Var(X + Y + Z) = 140/25
PXY = -1/2
[Hint: See Problem 22(a) for computtng Cov(X, Z ) . ]
(b) Var(X + Y) = 5/9, (c) No
25. E(X) = 14/5, Var(X) = 14/25, E(X) = 14/5, Var(X) = 14/25n
(d) Var(X + Y) * Var(X) + Var(Y) sLnce Cov(X, Y) * 0.
26. (a) fix = -1/4, ax = \T/4
6. E ( W , ) = 6 5 , V a r ( W , ) = 425/4
(b) There are 3Z possíble samples; the p . f . of X will be:
E(W 2 ) = 200/3, Var(W 2 ) = 400/3
x -1 -1/2 O 1/2 1
7. E ( W , ) = 65, V a r ( W , ) = 125/4
E(W 2 ) = 200/3, Var(W 2 ) = 200/3 f(x)' (3/8) ¿ 2(3/8) (1/2) 11/32 2(1/2) (1/8) (1/8) 2

A negativa covariance indícales that X, and X 2 are inversely related. (c) E(X) = -1/4, Var(X) = 7/32
The negative covariance has reduced the V a r ( W , ) and V a r ( W 2 ) .
27. (b, O,
8. E ( W , ) = 6 5 , Var(Wj) = 325/4
E(W 2 ) = 200/3, Var(W 2 ) = 1000/9 28. (a) E(W)=/j, (b) Var(W) =
When Cov(X, , X 2 ) = O, X, and X z are onrelated.
10. E(T) = nny., Var(T) = na2 (c) a = (d) a = 1/2

11. (a) 110, 325, -10, 325, (b) O, O, 2, 2 J_

(a> 10' 90' E(T)=1' V
304 Answers Answers 305

(b) .1 .2 .3 .4 .5
Chapter 9
1. (a) .0170, .0989, .6553; (b) O, .6481, .0138, (o) .8 Pr(T = 0) .5905 .3277 .1681 .0778 .0313
2. n = 10, p = .6 (e) The distribution of T can no longer be approximated by j binomial
(a) .9877, (b) .8205, (c) .9537
since a sample of 150 from 500 is a 30% sample. The random
variable T will have a hypergeometric distribution.
3. 69
17. Pr(engine does not fail) = 1 - q and Pr(engine fails) = q; henee,
4. (a) 7/27, (b) 26/27, (c) 2/9 Pr(two-engine plañe runs) ~ 1 - q 2 and Pr(four-engine plañe runs) =
1 - q 4 - 4q 3 (l - q). The two-englne plañe will be preferable when
6. (a) np, np(l-p), (b) S ~ b(s¡ n, p), (c) .0019
1 - q 2 > 1 - q4 - 4q 3 (l - q) or q 2 (3q - l)(q - 1) < 0. If q = O, 1/3, 1,
7. (a) P = S/n, (b) p, p(l - p)/n, (c) .0019 the performance will be equal. For O < q < 1/3, four-engine plañe pre-
(d) As n becomes larger, Var(P) gets smaller; that is, P will approach ferred and for 1/3 < q < 1, two-engine plañe preferred. Since q would
usually be much less than 1/3, the four-engine plañe would more than
Its true valué p.
Hkely be preferred.
8. n(l - p), n ( l - p)p
20. (a) 3/4, (b) E(X) = l/3, Var(X) = 4/9
9. p = x/n
21. (a) k = 1 - r, (b) E(Z) = Var(Z) =
12. Pr(Felíx gets three healthy bírds on a single try) = 1/10; 1 - r' (1 - r) 2
Pr(two successes in"three tries) = .027.
23 . i a) ——
13. (a) .8542, (b) .7463
14. (a) Let X be the number of uncooked peas in a sample of 10 peas. 24. (a) E(X) = 2 ; expected cost = $200.
Since the sample size is small in comparison with the population (b) Máximum probability occurs when X = 1 or X = 2; henee, most
size of rwo pounds of peas, X is approximately b(x; 10, p). probable amount of money is $150.
Pr(declaring peas cooked) equals Pr(X = 0) = (1 - p ) 1 0 . Henee (c) .0526
.1 .3 .5 .6 >. 6 25. X ~ p(x; 2); henee Pr(X < 5) = .9835 and Pr(X < 4) =.9473. Thus the
merchant should have at least five Ítems.
Pr(X = 0) .3487 .1074 .0282 .0060 .0010 .0001 .0000
27. Probability of error = .02
(b) .3487
(a)Let X¡ = number of errors Ln a block of 10 words; X j ~ b(x¡¡ 10, .02).
15. (a) Here again X is approximately b(x; 10, p). Now, however, 10
Then T = £.= 1 x. ~ b(t; 100, . 0 2 ) . SLnce n is large and p is small,
Pr(declaríng the peas cooked) is Pr(X = 0 or X = 1) = P*. Henee approx.
T ~ p(t; 2) with Pr(T < 1) s .4060.
.1 .3 .4 .5 .6 • .7 (b) T ~ b(t; 100, .02) which again can be approximated by a Poisson
with X - 2. The probabilíty is the same as in (a).
P* .7361 .3758 .1493 .0463 .0108 .0017 .0001 .0000
28. Let X¡ = number of students opposing the new grading scheme ín a
Note that with this sampltng scheme for all valúes of p the proba - sample (WOR) of n¡ from ith class for i = 1, 2, 3, 4. Then X. - h(x-;
bility of saying that the peas are cotnpletely cooked is higher than approx.
when the pan is accepted as fully cooked only when there are no ¿, N¡, n¡) or b(x.; n¡, . 1) since n¡ is less than
uncooked peas in the sample. sample for all i. Thus the total number opposing is T = 2 X. approx.
(b) .7361 approx.
b(t; £ n¡, .1) or T b(t; 100, . 1). Since Z n, is large and p is
16. (a) T is actually a hypergeometric variable, but its distribution can
be approximated by a bLnomial distribution since the sample size 5
is small in comparison with the population size.
306 Answers Answers 307

approx.
ama.ll, the distribution of T can further be approximated by T 2. (a) H 0 : p = . 2 5 vs. HA: p > .25.
p(t¡ 10) with Pr(T > 10) = .5421. (b) Prob-value = Pr(X >_ 6 I X ~ b(x; 10, .25) = .0197. For a = .05
we would reject H0 and conclude that the new treatment t is better.
29. Let X. = number of defectives Ln sample of 100 from eacb machine for Since the prob-value is greater than a, we would fail to reject H 0 .
approx. approx.
i = A and B; X p(X¡ 1), X, p(x; 2 ) . Henee
3. (a) H 0 : p = .6 vs. H A : p< .6.
X B ~ P(s¡ 3) and Pr(S > 1) = . 8008.
(b) Prob-value = Pr(X £ 3 I X ~ b(x; 10, .6) = .0548. For a = . 05 we
25! would f a i l to reject H0 .
31. (a)
17! 6 ! 2! .6496, .3922, . 1719. These probabilitíea indícate that the event
X <^3 is far more probable than when p is ,6; henee, one would
(b) 12.5, 10, 2.5
conclude that ¡t is more reasonable to reject the claim made by A
32. (a) Let X = nomber of donors with a particular blood type; and conclude that p is actually less than .6.
X ~ b(x; 5, . 0 3 ) .
Pr(8 < X £ 15 j X~ b(x; 15, .7) = . 9 4 9 9 ; henee, a, the probabüity
(b) Let Y = number of defectives in sample of 25 seta; Y ~ h(y; 4 0 0 ,
of rejecting the nuil hypothesis when it is true, for this critical
5000, 25). SLnce 5000 » 25, Y app-íox' b(y; 25, . 0 8 ) , whlch, since región is . 1760. This is too large a valué for a. In order to de-
n = 25 is reasonably large and p = .08 ¡s small, can be approxi- crease a, the valúes for which H0 ¡s accepted sbould be increased.
mated by a Poisson with \ 2. .4967
Let T = number of fish caught; t ~ nb(t; 10, .3).
1/2' 4 , (b) 1/2U
Let X = n u m b e r in sample- produced by first shift; 'X ~ h(x¡ 400,
If the methods are equally effective, tben the ranks of the children
1000, 5) or X aPP~rOX' b(x; 5, . 4 ) .
for method I! represent a random selectíon of three numbers from
(e) Let Z = number of cuatomers with colored telephones; Z - h<z; 100,
annrox. approx. 1, 2, 3, 4, 5, 6. Each selection has probabüíty l/( j = 1/20.
1 0 0 , 0 0 0 , 100) or Z ~ b(z; 100, .001) or Z p(z; - 1 ) .
The sample space consists of 20 points:
33. (a) Pr(X = k) = (.9) (.1), k = 1, 2,
(b) Pr(X > t) = (.9)*
15
(c) Pr(X = j I X > t) = (.!)(.9)J-1-1 , for j = t + 1, t + 2, . . (4, 5, 6)
(3, 5, 6) 14
34. (b) 1/6, (c) =(|)
(2, 5, 6) (3, 4, 6) 13
2 x-2 (1, 5, 6) (2, 4, 6) (3, 4, 5) 12
35. (a) f(x) = ( X ) (r) (7) for x = 2, 3,
(1, 4, 6) (2, 3, 6)(2, 4, 5) 11
(c) E(X) = 8/3, Var(X) = 8/9 (1, 3, 6)(1, 4, 5) (2, 3, 5) 10

(1, 2, 6)(1, 3, 5)(2, 3, 4) 9

36. (a) (í) ("Xo) , (") 1 - Z (* 7 <b>
d, 2, 5)(1, 3, 4) 8
(1, 2, 4) 7
Chapter 10
(1, 2, 3) 6
1 (a) H 0 : drug has no effect va. H A : drug slows response rate
t 6 7 8 9 10 11 12 13 14 15
(b) p = .5
(c) X ~ b(x; 10, .5) 1 1 2 3 3 3 3 2 1 1
(d) Prob-value = Pr(X >_ 9 I X ~ b(x¡ 10, . 5) = . 0108; henee H0 would f(t) 20 20 20 20 20 20 20 20 20 20
be rejected for any a greater than or equal to .0108. If H<, ís rejected,
(b) E ( T ) = 1 0 . 5 , Var(T) = 5.25
we conclude that the drug slows the reaction time.
 Answers 309
308 Answers

H : no improvement vs. H : ¡mprovement

(c) If method II ¡s more effective, T < 1 0 . 5 (¡.e., smaller valúes being O A
more convincing). The observed event T = 7 seems to indícate that
is equivalent to
method II (not I) is more effective.
H : p - .5 vs. H A : p > . 5 .
(a) Let U indícate up and D indícate down. Then a'typícal point ín S
is (Xi , x 2 , x 3 , x 4 ) with x¡ = D or U. There are 2* = 16 points in S. (c) Y ~ b(y; 8, .5)
Let C = n u m b e r of changes (d) .0351
(e) Reject H0 and conclude that the treatment ¡s effective In ímproving
E. C the breathing capacity.

(UUUU)(DDDD) 10. (a) Here X is counting the number of occurrences of the rare trait in
a randotn sample of 1000 peoplei X ~ b(x¡ 1000, .01). Since n is
(UUUD (DUUU)(UUDD)'| large and p is relatively small, we can approximate the distributíon
(DDUU)(DDDU) (UDDD)) of X by the Poísson with A = 10. Then

(UUDU) (UDUU) (UDDU) ) H : \ 10 vs- H :< 10.

O A
(DUUD)(DDUD)(DUDD) j (b) Prob-value = Pr(X £ 4 I X ~ P(x; 10) = .0293. For a = .05, we
(DUDU)(UDUD) would reject H0 and conclude that the rate is less than one in 100
people. Note that here the genetic trait is not nearly so rare as
c O 1 in Example 4 of this chapter; however, the prob-value Is exactly
the same sínce the observed valúes of X and p are the same ín
Pr(C = c) 2/16 6/16 6/16 2/16 both problems.
Pr(C > 2) = 1/2 11. (a) H 0 : water ts not polluted vs. H^: water ¡s polluted, which is
equivalent to HQ: p = . 1 vs. H^: p > . 1-
(c) Let X = number of cases with two or more changes; then X ~ b(x;
n , 1/2). Assumption: (i) Pr(two or more changes) is same for each (b) Prob-value = Pr(X > 4 I X ~ b(x; 20, . 1) = . 1330; therefore, do
case, ( i i ) Cases are independen!, ( i i i ) Considering only two out- not reject H0 for a = . 0 5 .
comes: > 2 and < 2 changes. (iv) Number of cases for consideration 12. (a) H Q : expect four will be c u r e d vs. H¿: m o r e than four will be cured
is fíxed ín advance. or H Q : p = .02 vs. H A ' p > - 0 2 .
(d) P r { X >_ 13 | X ~ b(x; 20, 1/2)} = . 1316. Since Pr(X > 13) is rea-
sonably large, the data support the hypothesls of randomness. (b) X ~ b(x; 200, .02); X is approximately P(x; 4).

Here we assume that the number of Ítems beíng produced by the assem- (c) Using the Poisson approximation, the prob-value is Pr(X >_ 6) =
bly Une is much larger than the 10 that have been sampled; henee, we .2148.
can assume that X, the n u m b e r of type A ítems in the sample, is approx- (d) We do not reject H0 and conclude that the nonsurgícal techníque is
imately b(x; 10, . 5 ) . The prob-value ¡s Pr(X < 2) = . 0547. If we take not as auccessful as aurgery.
a to be .05, we would fail to reject the claim that the two types of 13. (a) No. of (P) O 1 2 3 4 5 6 7 8
ítems are being produced in equal frequency. Here we would have ex-
pected f i v e type A ítems ¡f the claim is true. It may seem somewhat O¡ O O 5 14 21 29 20 9 2
strange that even when we observed only trwo type A ítems we were not E¡lp=.6 , 0 7 . 7 9 4 . 1 3 12.39 2 3 . 2 2 2 7 . 8 6 2 0 . 9 0 8.96 1.68
able to reject the clalm. This is due to the small sample size. In this
problem w i t h a = .05 we would have been able to reject the claim only E ¡ l p = .7 .01 .12 1.00 4 . 6 7 13.61 2 5 . 4 1 2 9 . 6 5 19.77 5 . 7 6
if we had observed either O or 1 type A ítem. In order to use the approximation for the goodness of fít statistic,
(b) If the treatment does not improve the breathing capacity, then p is we have to ensure that labeled ' O 1 , '!' and ' 2 ' need to be collapsed.
equal to . 5. Then
310 Answers Answers 311

E f f e c t i v e l y , there are then only six celjs and the v a l u é of the sta- p . The possible 2 x 2 tables which might have resulted from this
control
t i s t i c is . 5680. experiment are
(b) Again collapse the cells as Ln (a). Now the valué of the statistic ¡s
4 7 , 8 8 . Clearly the model with p = . 6 fits much better than that
with p = . 7 .
14. X ¡s the n u m b e r of red balls in the sample Using the hypergeometric distribution, the probabilities corresponding
x 0 1 2 3 to the tables are {1/252, 25/252, 100/252, 100/252, 25/252, 1/252}.
The prob-value is 25/252 + 1/252 or 26/252 = .1032; henee for a = .05
0| we would f a i l to reject the nuil hypothesis and conclude that the treat-
E¡ I WR ment is not superior.
E¡ I WOR Chapter 11
The goodness of fit statistic for WR is 4 . 7 1 with a prob-value of approx-
1. (a) The density function is a rectangle with a base of two units, ex-
imately . 2 and for WOR the statistic is .4349 with a prob-value of more
tending from -1 to 1, and a height of 1/2.
than . 5 . Based on these prob-valuea we cannot reject either model;
(b) Since the base of the rectangle is 2 units and the height is 1/2, its
however, the WOR raodel does appear to fit better.
área will be 1.
15. For the case in which the tetrahedron is rolled 80 times, the valué of (C) 5/8
the goodness of fit statistic is 9 . 9 . Here there are four classes. For (d) X B = -1/2 and x,0 = 4/5
a = .05, the model would be rejected and we would conclude that the
2. (a) The density function is a right-angled trlarjgle with a base of two
tetrahedron is not balanced.
units and a height of one.
16. X is the number of accidents. (b) 5/16
(c) x ¿5 = 1 and x,0 = 6AÍ10
O 1 2 3 3. (a) Form of the density function ia the same as in Problem 2; now the
base extends from -1 to 1.
80 61 13 1 (c) 45/64
E¡ U = .5 94.01 47.01 11.75 2.23 (d) x zs = O and x,0 = 6A/TÓ - 1 = . 8974
4 . c = l/(b - a). Note that the p . d . f . in Problem 1 is a special case of
this distribution.
For \ . 5 the goodness of fit statistic is 7 • 0626 and for X = • 6 it is
4 . 4 6 0 9 . Agaüi the prob-value Is greater than .05 for each of these 5. (a) The density function is an isósceles triangle with a base from -1
modela; however, it ís reasonable to assume that the distribution with to 1 and a height of 1.
A = .6 fits better. (c) Median - 0. Since the distribution is symmetric, its mean is equal
to its median.
17. Here the data should be arranged in a 2 x 2 table:
6. (a) .413, (b) .800, (c) .054, (d) .005, (e) .946
Treatment Control 7. (a) 1.645, (b) -.46, (c) 1.125, (d) 1.63
Alive 4 1 5 8. 7.5065
Dead 1 4 5 9. (a) - . 6 7 5 , (b) 203.125
10. P r ( a n y one individual has a craníal capacity > 117.6) =
5 5 10
117.6 - 105 \z
15
Here we are t e s t m g H: p. ^ =p , vs. H .: p. Let Y be the number in the random sample who have a cranial capacity
O treatment control A treatment
>117.6. Then Y ~ b(y; 10, .2) and Pr(Y > 3) = .3222.
312 Answers Answers 313

C h a p t e r 12 / 8 9 . 5 - 80\. (b) X is approxLm

1. (a) n = 3.5, ti2 = 35/12 .009.

(c) Reject the nuü hypothesis and conclude that the new medfcation is
(b) M = 7, cr2 = 5.8
superior to the standard.
(c) M = 1.5, o-2 = .75
17. (a) Y is approximately N ( 2 0 , 10)
The table below gives the Pr{ |X - ^\ ccr} for these three 27. 20 \ j = P r ( Z > 2 . 3 7 ) = .009.
(b) Prob-value is P r ^ Z >
distributions:
(c) Here the nuil hypothesis would be rejected for any valué of a less
(a) (b) (c)
than .01.
/ 14.5 - 10 \. (b) X is approx
c = 1.5 1 .833 .75
c = 2.0 1 .944 1 P r ( Z > 1.5) = . 0 6 7 . For a = .05, we would fail to reject the nuil
hypothesis.
c = 2.5 1 1 1
19. Here X, the number of successes with the new method, has a hypergeo-
c = 3.0 1 1 1
metric distribution w i t h M = 50, N = 100, N - M = 50, n = 50; henee,
X is approximately N(25, 6 . 3 1 ) . Then the prob-value ís
In each case these probabilities should be compared with 1 - — / qq c 9S \z > ' ) = 1 -4(5.77) = 1 - 1 = 0. This leads to rejection

3. When O < c < 1, then —j > 1. The theorem is not useful because it
— c2 for all a. Note that the proportion of successes is the same as in Prob-
would give probability bounds greater than 1. lem 17 in Chapter 10. Here, however, the larger sample size allows
. us to confirm the superiority of the new method with high credibility.
4 . (a) E(X) = O and Var(X) = 2k 2 p
(b) -k 0

f(x)

5. (a) < . 25, (b) > .75

k = '•/SO
8. n = 32
12. and 13.
(a) (b) (d)

Exact binomial
probability .1536 .4096 .9728 .1808 1.0000
Approx. normal
probability .173 .300 .932 .198 .948
14. The normal approxünation is not adequate since n = 4 is too small.
15. Here X ~ b(x; 120, and X is approxímately ~ N(96, 19.2). Using
the normal approximation, we get
(a) .069, (b) .026, (c) .060