APPLICATION OF SAMPLING METHODS IN BIOMEDICAL STUDIES 21If the persons to be sampled live in a very small area, it is possible to drive or
walk around the area and note on a map all the places that people live. This is more
difficult than it sounds, as people can live in unusual places such as garages or trailers
or in locked residential areas where access to the households is virtually impossible.
Sometimes a reverse directory, supplemented by a tax roll or aerial photographs, is
helpful. Then a sample of households can be taken. Often, a systematic sample is the
simplest to use. From a sample of households, a subsample of adults can be obtained.
Another method of sampling households is to do a random-digit dialing survey by
telephone. The first step is to obtain the area code (3 digits) and the prefix numbers
(3 digits) in the desired area. Other information on how the last 4 digits are assigned
by the telephone company can be obtained from the company. The next step is to dial
4-digit random numbers with suitable prefix numbers (see Frey [1989] for a more
detailed description of how to take a sample by telephone). Sometimes investigators
will take random samples from telephone directories, but this procedure results in
many households not being sampled because they have an unlisted number or a cell
phone. Note that persons without telephones have no chance of being surveyed.
As the population gets smaller and the eligibility restrictions for defining who is
in the population get stricter, the usual problem is obtaining a sufficient sample size.
In many biomedical studies, this is a real concern. For example, if you wish to study
persons who just came down with a rare disease, you will tend to take anyone you
can find who meets the entry criteria. This type of sample is often called a chunk or
convenient sample. It will not make sense to a physician to go to a lot of effort to find
suitable subjects and then decide not to use some of them. Such samples do not meet
the criteria of measurability discussed earlier, but they may be the best available. For
example, it is difficult to know if patients with a particular disease who are treated by
one physician are typical of all patients with the same disease.
In public health surveys, investigators often want to sample persons who follow
procedures or lifestyles that are illegal or unpopular and hence hard to find. Samples
of volunteers can be recruited using suitable newspapers, television, radio, or Internet
websites that are aimed at the desired population. For example, a sample of heavy
marijuana smokers cannot be recruited in the same manner as the general public is
recruited. The resulting samples are often chunk samples, but they are the only type
that is feasible and economical.
Another method of doing surveys is to use the Internet. If one has the e-mail
addresses of a group of people that they want to survey, web-based outfits like Sur-
veyMonkey can be used (see http://www.surveymonkey.com/)..)
2.4.3 Samples for Experiments
In laboratory experiments, where animals, tissue cultures, or other observational units
are used, the sampling method most used is chunk samples. Investigators tend to be
careful where they purchase their animals and the type used for a given kind of
experiment, but the focus is on how the treatment affects the outcome and how best to
measure that outcome. An implicit assumption is made that the effects of the treatment
are quite general and the animals or whatever is being experimented on can be thought