176 Solutions to Selected Exercises
included. When the investigator and/or the patient know which treatment group
they are in before the completion of the treatment, they could act in a way that
creates bias in the estimates. If both the investigator and the patient are unaware of
the treatment the patients are more likely to all be treated in the same manner and
bias will not creep into the study.
Chapter 2
- Describe and contrast the following types of sampling designs. Also state when
if ever it is appropriate to use the particular designs:
(a) Simple random sample
(b) Stratifi ed random sample
(c) Convenience sample
(d) Systematic sample
(e) Cluster sample
(f) Bootstrap sample
(a) Simple random sampling is just sampling at random without replacement from
a well - defi ned population.
(b) Stratifi ed random sampling is a sampling procedure where the data are divided
into groups (strata) that make the subpopulations homogeneous groups. In each
strata, a specifi c number patients are sampled at random without replacement.
So it is a collection of simple random samples drawn for each strata. Stratifi ed
random sampling is better than simple random sampling when subpopulations
are homogeneous, and there are differences between the groups. If the original
population is already very homogeneous, there is no benefi t to stratifi cation over
simple random sampling. It is possible to obtain unbiased estimates of the popu-
lation mean by either sampling technique, but one estimate will have a lower
variance compared with the other depending on the degree of homogeneity
within and between the subpopulations.
(c) A convenience sample is any sample that is collected in an operationally con-
venient way. This is usually not an acceptable way to sample because it is not
possible to draw inferences about the population from the sample. This is
because inference depends on having known probabilities for drawing elements
from the population.
(d) Systematic sampling is an ordered way of selecting elements from the popula-
tion. So, for example, if you wish to take a 20% sample, you can enumerate
the population and draw the fi rst and skip the next four until you have run
through the entire population. Systematic sampling can sometimes be easier
than random sampling, and if there is no pattern to the ordering it may behave
like a simple random sample. However, if there are patterns, such as cycles,
the method can be extremely biased. In the 20% sample, suppose that the data
bboth.indd 176oth.indd 176 6 6/15/2011 4:08:21 PM/ 15 / 2011 4 : 08 : 21 PM