Basic Statistics

PROBLEMS 25

privacy reasons. A list is made of 10-15 nearby houses following an agreed-upon
pattern. The interviewer visits the first household on the list and sees if an eligible
control lives in the house (say, of the same gender and within 5years of age). If
the answer is no, the interviewer goes to the next house, and so on. Letters are left
when no one is home, in an attempt to encourage a contact. Not-at-home addresses
should be contacted repeatedly since the investigator does not want a sample solely
of stay-at-homes. Neighborhood controls are costly and time consuming to obtain,
but they are thought to be suitable controls.
It is difficult to describe the sample of controls in a straightforward way since they
are chosen to match the cases. Viewed as a single sample, they certainly lack mea-
surability. Their value lies in their relationship to the cases. For further discussion of
case/control studies and how to choose controls, see Rothman [ 19861 or Schlesselman
[1982].
As can be seen from this introduction to sampling for medical studies, the problems
of feasibility and economy often lead to the use of samples that are not simple random
samples. It takes careful planning to obtain samples that, if not ideal, can still be used
with some level of confidence. The success of a good study rests on the base of a
good sample.

PROBLEMS

2.1

2.2

2.3

2.4

2.5

2.6

Draw three simple random samples, each of size 6, from the population of blood

cholesterol measurements in Table 2.1. Use the table of random digits to select

the samples, and record the page, column, and row of the starting digit you use.

Record the six individual measurements separately for each sample in a table.

They will be used for subsequent problems in other chapters.

Draw one systematic sample of size 6 from the population of blood cholesterol

measurements. Record what your starting value was.

Suppose that you decide to use Table A. 1 to assign patients to treatment A or

B by assigning the patient to A if an even number (e) is chosen and B if odd

number (0) is chosen. Starting with the numbers in the upper left-hand corner

of the first page of the table and reading across the first row, you would have

oe eo 00 eo oo-oe oe eo 00 ee-oe eo 00 ee oe-ee 00 oe 00 00-00 eo eo

eo eo. Does this appear to be a random allocation to you? How do you explain

the fact that six odd numbers were chosen in a row? If you use this allocation,

will you have an equal number of patients receiving treatments A and B?

You are asked to take a sample of 25 students from a freshman medical class

of 180 students in order to find out what specialty they wish to enter. Write

out instructions for taking this sample that can be followed by a staffer with no

sampling experience.

Use a random number generator from a statistical computer package to assign

20 patients to two treatments, with 10 patients receiving each treatment.

An investigator wishes to take a sample of 160 undergraduate students from

a university to assess the attitudes of undergraduates concerning the student

health service. The investigator suspects that the attitudes may be different for