15 November 2000
Test 2 Answer Guide
Question 1. The four methods (aside from "rigged"!) are random,
stratified random, cluster, and systematic. See 17 October class
notes and the beginning of Chapter 7.
Question 2. The null hypothesis H0 is "nothing much
going on" -- µ = 1. The alternative hypothesis Ha is
µ < 1. A Type I error (false rejection of H0)
would lead to
a false accusation of the supermarket. A Type II error (false acceptance
of H0) would result in customers being cheated. See the
early part of Chapter 9 and 31 October class notes.
Question 3. Attack: a biased estimator may, if the bias
is reasonably small, be superior to an unbiased estimator with very large
variance. See Section 8.2 and 24 October class notes.
Question 4. Continuous and bounded both ends: uniform, triangular,
beta. Continuous unbounded on both ends: normal, Student's-t.
Continuous bounded on exactly one end: lognormal (logarithms can't
be negative!), chi-square, F, gamma, Weibull, exponential.
Question 5. Attack: There may be a strong non-linear
relationship. See Section 6.5 and 5 October class notes.
Question 6. This question is similar to assigned homework from
chapter 6. Using a double integral, c = 3 to keep the probability
1. Similarly, the probability that 0 <= x <= 0.25 and
simultaneously
0 <= y <= 0.5 is 0.022. Since the product of the marginal
densities
in x and y equals the density function, x and y
are independent by definition 6.9 page 274.
Question 7.
For group X, 115 has z-score = 3.0. For group Y,
115 has z-score 2.4. So 0.0013 is the fraction of group
X
qualifying and 0.0082 is the fraction of group Y qualifying.
So the expected fraction of the workers from group X is
0.0013/(0.0013+0.0082)
or less than one-seventh and the expected fraction from group Y is
greater
than six-sevenths. Shocked? This problem has much to say to
those who would whine "Our test may be culturally or racially biased, but
only a little!"
Question 8.
Use the chi-square test from section 8.10. See assigned homework
problem 8.60.
Question 9.
Use the t-test for small samples (box page 366). See assigned
homework problem 8.28.
Question 10.
Use the new interval construction method presented in
class and illustrated with exercise 8.44 on 31 October. You
presumably used this method to solve exercise 8.48.
Question 11.
This question is similar to exercise 7.24 page 318, and uses the central
limit theorem. The breaking strength of the cable is normally distributed
with mean 15,000. From box page 236, the variance of the strength of each
wire is 450. The variance of the strength of the cable is 45,000, so the
pertinent standard deviation is 212.1. The pertinent z-score is
(14,705 - 15000)/212.1 = -1.39. The chance a standard normal
z is less than this is about 0.08, which is the probability of breakage.
Question 12.
This question is similar to exercise 9.4 page 429. Use table 3 page
1092, as you did for Test 1 problem 10 or homework exercise 4.70. If
l = 4.0, the chance of 7 or more such
storms = 1 - 0.8893 = 0.1107, causing an incorrect rejection of
H0
(Type I error). Hence a = 0.1107. If l
really is 5.0, the chance of 6 or fewer such storms, causing an incorrect
acceptance of H0 (Type II error), is 0.7622. Hence b
= 0.7622.