Assignment 6 Economics 31 Fall 1999

Confidence Intervals and Hypothesis Tests

Reading: Mirer Chtr 18, Beals Chtr.8 (for additional help - read ahead

Mirer Chtr. 11,12,13, Beals Chtr. 9,10,11)

1. A small internet start up company is considering buying adspace on a
major newspaper's website. Before purchasing this ad, however, the company
wants an estimate of the average yearly computer software purchases of the
households patronizing this newspaper's website. The company contracts with
a market survey firm to randomly sample 100 family heads whom own their own
computers and patronize this website daily and estimate, at the 95% level
of confidence, the mean expenditure in the population. The contract with
the market survey firm provides that the start up company will pay $25 per
completed interview. Assume that the population distribution of
expenditures is normal and that the population standard deviation is known
to be $68. Suppose the report from the market survey firm states that the
95% confidence interval has a lower limit of $56.15 and an upper limit of
$93.85. The report is accompanied by a bill for 100 interviews. Explain why
the above confidence limits might lead one to suspect that the market
survey firm did not complete 100 interviews. How many interviews were
actually made?

2. A certain multinational tobacco firm is considering replacing the type
of tobacco used in one of its namesake brands from blended American style,
to an all-Bright, Virginia tobacco, labeled as stronger and less sweet.
(Richard Kluger 116 Ashes to Ashes) They sampled 200 randomly selected
consumers who regularly smoke cigarettes and gave each of them one box of
each type in a blind taste test. The result was that 160 preferred the
blended American style. Construct a confidence interval at the 95% level of
confidence for the population proportion of cigarette consumers who prefer
American style.

3. The Gallup Poll periodically takes a random sample of about 1500|
Americans. The percentage that favor the acceptance (decriminalization) of
marijuana possession increased from 23% in 1979 to 29% in 1999.

a. Construct a 95% confidence interval for the population percentage in
favor, each year

b. Find a 95% confidence interval for the change in this percentage from
1979 to 1999.

4. To see what difference class attendance made, Professor Hollister
sampled grades from his large statistics class of 530 students. From the
220 students who attended class less than half the time, he took a random
sample of 5 grades. From the remaining 310 students who attended at least
half the time, he took an independent random sample of 5 other grades:

irregulars regulars

41 69

81 56

52 83

69 70

62 92

a. Construct a 95% confidence interval for the mean difference between the
two groups of students.

b. Explain, without necessarily doing the calculations, whether the
interval would be wider or narrower under each of the following conditions:

i. each sample contained thirty students.

ii. an 80% confidence interval was required, with everything else the same
as in part a.

c. To what extent does part a support the contention that ''it is worth 18
points to come to class regularly''?

5. In a random sample of 100 Democrats, 27 were in favor of marijuana|
legalization. Another independent sample of 100 Republicans, found 22 in
favor of drug legalization. The confidence interval .02<Px - Py<.08 was
calculated for the difference between the population proportions (for more
info see http://www.gallup.com). What is the probability content of this
interval?

Part 2

1. A professional photographer uses a battery pack to power his strobe
lights when he is doing a photo session. He has purchased a new type of
battery pack that is supposed to average 100 flashes before it needs
recharging. He has recharged the pack 10 times. The mean number of flashes
was 115 with a simple standard deviation of 8. Assume that the distribution
of flashes per charge is normal. Test at the 5% significance level the
hypothesis that the mean number of flashes per charge is equal to 100.

2. The Economic Policy Institute published data in 1999 suggesting that
17.8 % of all households in America are headed by single females (Cornell
University).Out of 300 households in Philadelphia randomly sampled by a
statistician, the number of households headed by single females is 45. Test
at the 5% significance level the hypothesis that the true proportion of
houses headed by single females is equal to 0.18.

3. There are two kinds of avocado, black and green; Person A wants to buy the
type that has the lowest cost per ounce. However, the vendor quotes her a
price on a per avocado basis. Suppose green avocados had a known standard
deviation of 2 ounces and black avocados had a known standard deviation of
3 ounces. If the mean weight of the green avocados is more than 5 ounces
above the mean weight of the black avocados, it will be cheaper to buy the
green ones. Person A takes a sample of both types. The sample mean of 30
green avocados is 22 ounces. The sample mean of 20 black avocados is 16
ounces. At the 5% significance level, can she reject the null hypothesis
that the mean difference is equal to 5 ounces? Assume that both population
distributions are normal.

4. Leo Burnett is in charge of a new promotional campaign for Philip
Morris. Philip Morris introduced a new package in 1954. According to
Burnett's agency, ''the lowercase m at the beginning of the Marlboro logo
served to trivialize the product. Similarly, the diagonal peppermint
stripes that formed the chevrons on the red ''roof'' of the pack were too
fussy distract[ing] from the clean, strong basic design'' appealing to
youth and adolescents (Kluger 181). Philip Morris's head, Joseph Cullman,
disputes this analysis and orders a test. To test the effect of the new
package, the company conducted a test in which the 100 outlets in a chain
of tobacco stores were divided into two groups of equal size. The first
group retained the old package and the second group used the new package.
The shelf space for each design was the same at every outlet. The average
sale per store during the test period was 350 cases for the old design and
360 for the new design, with sample standard deviations of 35 and 40,
respectively. If the 100 outlets were considered a representative sample of
all stores that carry Marlboro's, should the hypothesis that the new design
will outsell the old design be rejected at the 5% significance level? Why
might the observed gain of 10 cases per store be an unreliable predictor of
the potential benefit of the new design?

Return to Econ 31