Nonprobability sampling selects units nonrandomly. It is very common in the behavioral science research, for example, psychology freshmen or Internet users are asked to participate in a study. Moreover, a probability sample becomes a nonprobability sample if some of the selected persons are nonrandomly missing or drop out of the study. Nonprobability sampling is threatened by systematic errors that bias the study results. Three procedures are described to control for bias. First, representative sampling: a sample is selected that is judged to be representative for the population. Second, applying weighting procedures in the analysis of the data. Third, comparing study results across different subpopulations (e.g., males and females, younger and older participants, and so on). If study results are approximately the same across different groups, generalization of the study results is supported. However, generalization from a nonprobability sample to a population cannot be based on statistical theory. Therefore, probability sampling has to be preferred above nonprobability sampling.
KeywordsGeneralization Other- and self-selection Reference sample Representative sample Weighting
- Rosenthal, R., & Rosnow, R. L. (1975). The volunteer subject. New York, NY: Wiley.Google Scholar