EPIB 650: Biostatistics I

Spring 2011
Homework 4 Solution

Chapter 7

Problem 17
(a) P(Z > 2.60) = 0.005
(b) P(Z < 1.35) = 1 - P(Z > 1.35) = 1 - 0.089 = 0.911
(c) P(-1.70 ≤ Z ≤ 3.10) = P(Z > -1.70) - P(Z > 3.10)
= 1 - P(Z > 1.70) - P(Z > 3.10)
= 1 - 0.045 - 0.001
= 0.954

(d) z0.15 = 1.04

(e) z1−0.20 =
− z0.20 =
-0.84

Problem 19

(a) X ~ N (172.2, 29.82 )

130 − 172.2
P(X < 130) = P(Z < ) = P(Z < -1.42) = P(Z > 1.42) = 0.078
29.8

210 − 172.2
(b) P(X > 210) = P(Z > ) = P(Z > 1.27) = 0.102
29.8
(c) P(X < 130 or X > 210) = 0.078 + 0.102 = 0.180

Y ~ Bin(5, 0.180)

5
P(Y ≥ 1) = 1 - P(Y = 0) = 1 -   (0.180)0 (1 − 0.180)5 = 1 - 0.371 = 0.629
0

1
Problem 20
(a) The probability of correctly predicting coronary heart disease for a man who will develop it is
260 − 244
P(Xd ≥ 260) = P(Z ≥ ) = P(Z ≥ 0.31) = 0.378
51
(b) The probability of predicting heart disease for a man who will not develop it, or the
probability of a false positive, is
260 − 219
P(Xnd ≥ 260) = P(Z ≥ ) = P(Z ≥ 1.00) = 0.159
41
(c) The probability of failing to predict heart disease for a man who will develop it, or the
probability of a false negative, is
P(Xd < 260) = 1 - P(Xd ≥ 260) = 1 - 0.378 = 0.622
(d) If the cutoff point is lowered to 250 mg/100 ml, the probability of a false positive would
increase, while the probability of a false negative would decrease.
(e) Initial serum cholesterol level is not very useful for predicting coronary heart disease in this
population, because the probabilities of false positive and false negative errors are both very
high.

Chapter 8

Problem 10

The distribution of means of samples of size 40 has mean µ = 1.81( µ g / m3 ) and standard
σ 2.25
deviation = = 0.36( µ g / m3 ) , and, assuming that n = 40 is large enough, is
n 40
approximately normally distributed. The central limit theorem applies even though the
underlying population of measurements is skewed to the right.

Problem 13

(a) X ~ N (341, 792 )

300 − 341 400 − 341

P(300 ≤ X ≤ 400) = P( ≤ Z≤ )
79 79
= P(-0.52 ≤ Z ≤ 0.75)

2
 = P(Z > -0.52) - P(Z > 0.75)
= 1 - P(Z > 0.52) - P(Z > 0.75)
= 1 - 0.302 - 0.227
= 0.471

792
(b) X ~ N (341, )
5

300 − 341 400 − 341

P(300 ≤ X ≤ 400) = P( ≤ Z≤ )
79 / 5 79 / 5

= P(-1.16 ≤ Z ≤ 1.67)
= P(Z > -1.16) - P(Z > 1.67)
= 1 - P(Z > 1.16) - P(Z > 1.67)
= 1 - 0.123 - 0.047
= 0.830

792
(c) X ~ N (341, )
10

300 − 341 400 − 341

P(300 ≤ X ≤ 400) = P( ≤ Z≤ )
79 / 10 79 / 10

= P(-1.64 ≤ Z ≤ 2.36)
= P(Z > -1.64) - P(Z > 2.36)
= 1 - P(Z > 1.64) - P(Z > 2.36)
= 1 - 0.051 - 0.009
= 0.940

792
(d) X ~ N (341, )
10
a − 341 b − 341
P(a ≤ X ≤ b) = P( ≤ Z≤ ) = P(-1.96 ≤ Z ≤ 1.96) =0.95
79 / 10 79 / 10

79
a = 341 - 1.96 = 292
10

3
 79
b = 341 + 1.96 = 390
10

Therefore, the interval (292, 390) encloses 95% of the means of samples of size 10. This
symmetric interval is shorter than an asymmetric one.

Problem 15

462
X ~ N (211, )
25
195.9 − 211 226.1 − 211
P(195.9 ≤ X ≤ 226.1) = P( ≤ Z≤ )
46 / 25 46 / 25

= P(-1.64 ≤ Z ≤ 1.64)
= P(Z > -1.64) - P(Z > 1.64)
= 1 - 2 × P(Z > 1.64)
= 1 - 2 × 0.05
= 0.90