VIII. Sampling

Unlike quantitative studies, which emphasize the use of random or systematic samples, the methods described here employ primarily purposive sampling. Random sampling involves selecting units from a sampling frame that uses randomly computer generated numbers and allows one to generalize the results to the population from which the sample is drawn with measurable statistical precision. The main purpose of a random sample is to ensure that the sample is representative of the total population. Purposive sampling involves selecting individuals based on their level of cultural knowledge of a particular research topic or area. Although purposive sampling does not allow generalization with a known precision, every attempt should be made to ensure adequate coverage of important sub-populations (ethnic groups, poorer groups, etc.). For example, one might sample four to five respondents from different economic or ethnic groups.

The methods in this protocol involve two types of informants. Key informants are purposively selected and are interviewed several times (at least 3-5 times) in order to provide in-depth information about particular topics. Respondents may be either purposively or randomly selected and are usually interviewed only 1 or 2 times. Their interviews are intended to ensure coverage of information. The two types of informants are intended to provide a balance between coverage and depth of information collected about women's health.

How do we locate key informants and respondents? Often the selection of key informants is based on the recommendation of knowledgeable community members. Commonly, community leaders are first asked to recommend persons who would be "good to talk to ..." about the topic of interest. These initial referrals usually lead to secondary and even tertiary referrals. A key informant is a person with whom the interviewer builds a social relationship of communication over time. Every referred person is a potential key informant, and it should be clear after 1 or 2 interviews whether this person will serve as a true key informant.

Part of the decision about random or purposive selection of respondents depends on the availability of current lists of persons/households in the study community. If a reasonably current community list can be obtained, random sampling of respondents can be used to strengthen the representativeness of findings. Many of the groups using this protocol will have access to household and/or individual lists based on health records. When a community list is not available, a systematic sampling framework sample may be used, such as mapping the study community and interviewing persons in every fifth household.