A sample allows you to make generalizations about populations. You can think of a sample as a subset of a population, but that subset is only useful if it accurately represents the larger population. To ensure that your sample accurately represents your population you must clearly define the characteristics of your population, determine the required sample size, and choose the best method for selecting members of your sample from the larger population. For example, it might be too expensive to interview all 940 managers who work for your client, but we may be able to get just as valid information with a smaller sample, as long is it is chosen properly. If you were to select 100 to obtain results with a questionnaire, how would you select members of this sample?
Random Sampling
The basic characteristic of random
sampling is that all members of the population have an equal chance
of being selected as part of the sample. You generally use a table of
random numbers for drawing members of your random sample. For our
example you would assign each of the 940 managers a number (000-939).
You could then "blindly" start at some point on a table of random
numbers and work your way up or down the column selecting out
managers as you encountered each three digit number between 000 and
939. Random sampling does not guarantee that a sample will represent
the population. It does guarantee that any differences between the
sample and the population they have been drawn from are only a
function of chance and not a result of researcher bias. You will use
inferential statistics to estimate how much the sample likely differs
from the population.
Stratified Sampling
When populations consist of a number of
subgroups you use stratified sampling to ensure that each of these
subgroups is represented in the sample. In our example of 940
managers, we might want to ensure that our sample reflects the same
percentage of field managers, office managers, and account managers
as found in our population. Since our population is composed of 30
percent field managers, 50 percent office mangers, and 20 percent
account mangers, then so should our sample. To find your sample from
this population, you would first break up your population into these
three groups, and then randomly select the appropriate number of
members from each group.
Cluster Sampling
Many times it is not possible to list all
the members of a target population, assign them numbers, and then
randomly select a sample. Cluster sampling overcomes this problem by
selecting groups instead of individuals. For example; suppose you
wanted to find out what "real-world" examples of applied math appeal
most to adult learners enrolled in basic education programs. It would
not be realistic to assign a number to every member of this
population and then use a random number table to select an
appropriate number of sample members. A more realistic approach would
be to randomly select a sample of adult education centers, and then
survey all of the members of those randomly selected centers. These
centers would be the "clusters" that comprise the sample.
Systematic sampling
In systematic sampling you select your
sample by choosing every nth member from a list that includes all members of the
population. From our example of 940 managers, if you know that you
want 100 members in your sample you would select every ninth member
of your list. You start by randomly selecting one of the first nine
members of your list and then taking every ninth member from that
point on. Even though your choices are not independent, a systematic
sample can be considered a random sample if the list of the
population is randomly ordered.
On to sampling activity.
