Network Sampling

Network sampling is widely used when rare populations are of interest in survey research. Typically, sampling frames do not exist for rare populations [Page 507]because usually there is little information on the size and magnitude of the population. Two main methods can be employed in a survey with an unknown population: screening and salting. The first way is to screen for respondents of interest during the interview, and the second approach is to acquire sample units through officiai records or documents. Both approaches have shortcomings. Screening requires high costs. Salting entails difficulties with obtaining membership information, because officiai personnel records may be confidential. Network sampling is considered an alternative to the previous ways of estimating rare populations in which sampling frame is almost impossible to obtain.

Network sampling is also called snowball sampling or multiplicity sampling. This sampling technique is widely used to estimate populations such as the homeless, Korean War veterans, and patients with rare forms of cancer. Particularly, it has been found that network sampling was much more efficient than other conventional sampling methods for estimating the number of cancer patients. Most network samples have employed family members, relatives, and friends as informants; network informants report about all persons in their network; and sizes vary to degree from respondent to respondent.

Indeed, obtaining an initial sampling frame is a prerequisite for any network sampling method, and the quality of that frame is essential for the ultimate success of the method. Although family members or relatives are used as the network for many of the sampling frames, it need not be restricted to them, depending on the topic of a study. For instance, let us suppose that Korean Americans living in Michigan are the target population of a study. Network sampling using families, relatives, friends, and even casual associates may be useful for this case. In addition to blood kinship, membership lists can be used: Korean religion membership lists, Korean association lists, and so on. That is to say, using multiple sources included in the network sampling increases the network frame in quality and coverage.

To estimate the unbiased survey statistics, weighting is necessary for network sampling; the total eligible respondents of a particular network are weighted by the reciprocal of one over the total number of the particular network. Generally, interviewer costs are a primary concern for the network sampling. In this method, interviewers should meet with potential respondents who were identified by informants to see whether the respondents are eligible for a particular survey. This process increases interviewer time and costs, though both depend largely on the size of network. However, this sampling reduces screening costs.