Use purposeful sampling when collecting qualitative data

Purposeful sampling is appropriate for qualitative data collection, such as focus
group discussions or semistructured interviews. Purposeful, also known as
nonrandom, sampling is the selection of participants based on their knowledge,
perspective or other characteristics of interest (e.g., females or males, young or old,
very poor or better off).
  Remember that you cannot generalize qualitative data generated from purposeful
sampling to represent larger populations. You should use data from purposeful
sampling to understand more about the context or situation of the respondents only.
  Review the information needs and required comparisons stated in your analysis plan
to determine which type of purposeful sampling is best suited for your survey.
Common types of purposeful sampling
Best- and worst-case sampling compares communities or individuals who are
considered best or worst cases based on designated characteristics. For example,
best- and worst-case sampling could look at participating households that are most
vulnerable and least vulnerable in a given community to characterize vulnerability
in the community and identify target groups for future interventions. Another
example would be to compare communities that had the highest and lowest rates of
completion for a given project. Here, you would use best- and worst-case sampling
to highlight the factors contributing to these various rates of completion. Best- and
worst-case sampling is not useful to understand typical cases or the common
context.
Typical case sampling provides greater understanding of the general scenario by
studying typical cases, meaning those that are average or not markedly better or
worse than others, according to the characteristics that are of interest. It is important
to resist the temptation to select best-case communities and call them typical; this
would create a bias in the data and misrepresent the project.

Critical-case sampling selects a sample of individuals, households or communities
with particular characteristics, based on the idea that they are critical to
understanding a context or situation. Interviews with community leaders or focus
groups with widows are examples of critical-case samples. They are useful to
understand particular perspectives of key stakeholders or of members of vulnerable
groups.
Quota sampling is designed to interview or include participants with particular
characteristics in proportion in the sample population equal to their proportion in
the community. For example, if an estimated 30 percent of households in a
community are female-headed, quota sampling would stipulate that 30 percent of
respondents must be from female-headed households and 70 percent from maleheaded households.
  Avoid convenience samples. A convenience sample includes individuals who are
readily available to participate in the study. There is a high degree of bias involved in
this method. For example, choosing a sample of communities that is close to a main
road may be convenient, but it is likely to show markedly different results than a
sample of communities that is several hours away from the main road.
  Inform communities in advance if they will participate in the survey. It is important
togive adequate warning so that household members can plan to be available on a
certain day (and at a given time if you can be that specific). If you do not inform
communities in advance, many individuals may be in the fields working when the
interview teams arrive, for example. This leaves the teams to interview only
individuals who are not in the fields because they do not own land, have access to land
or have access to the required agricultural inputs and may bias the sample.