As we mentioned on the
previous page, two important characteristics we seek in our samples are
randomness and representativeness. Randomness ensures that
there are no biases in our sampling procedure and, if we know the important
distinguishing features of the population we are interested in, such as their
position in the company, geographic location, gender, or the like, then we can
test to see how representative the sample we obtain is of the group as a whole.
When there are imbalances, we can correct for these statistically. We can
weight the sample to bring it more into line with the proportions
of members of the sample in various subgroups in comparison to members of the
population in those subgroups.
When you are designing
your research program, take time to think about the important distinctions that
you are interested in among the group that you are seeking to learn more about.
Are all respondents equal in your eyes? If so, then a simple random sample
is likely to meet your needs well. There will be many cases, however, that
will prompt you to be interested in subdividing your sample when you are
interpreting and acting on the data. You may be interested in where your
employees or customers are located, what divisions they are in or what job
titles they hold, or how large a customer the company they represents is for
your business. In these cases, you may want to select more specific
samples using certain rules or procedures.
Often, this means defining
and drawing a stratified, random sample in which you define the subgroups
and then select from within them to help to ensure both that you are likely to
have enough people among your respondents to make accurate statements about the
subgroups themselves and that when grouped together they provide a
representative picture of the population as a whole. Although a random
sample is still the goal, stratifying the sample in one or more ways can help to
ensure that a complex population that interests you is well-represented within
the subgroups of the sample that you obtain data from. For multinational
companies, this may mean sampling to ensure sufficient numbers and
representation to look at the data by country or world region. For other
purposes, you may be interested in people within various positions or divisions,
in various age groups, or in members of other predefined subgroups within your
employee or customer base. Sophisticated sampling procedures can help to
ensure that you get the information you need to manage your business most
effectively.
