Non Probability Sampling

Non probability sampling is a sampling approach in which members of the population do not have a known, nonzero chance of being selected for the study. Selection is based on the researcher's judgment, convenience, or specific criteria rather than on random selection, and includes methods such as purposive sampling, convenience sampling, snowball sampling, and quota sampling.

While probability sampling (where every member of the population has a known, nonzero chance of selection) is the gold standard for quantitative research aiming to generalize results to a population, non probability sampling is widely used in both qualitative and quantitative research for practical and theoretical reasons. Qualitative research often uses purposive sampling to select participants who can provide the richest information about the phenomenon under study. Pilot studies and exploratory research use convenience samples when the goal is to test procedures rather than to estimate population parameters. Hard to reach populations (people who inject drugs, undocumented immigrants, rare disease patients) may be accessible only through snowball sampling. Understanding the strengths and limitations of non probability sampling is essential for designing feasible studies and interpreting their findings appropriately.

Several non probability sampling methods are commonly used. Purposive (or purposeful) sampling selects participants based on specific criteria relevant to the research question, such as expertise, experience, or key characteristics. Subtypes include maximum variation sampling (seeking diverse perspectives), homogeneous sampling (selecting similar participants to study a specific subgroup in depth), critical case sampling (selecting cases that are particularly informative), and theoretical sampling (selecting participants guided by emerging theory, as in grounded theory).

Convenience sampling selects participants based on availability and willingness. Students in a psychology class, patients at a single clinic, or shoppers at a mall are common examples. It is the easiest and cheapest sampling method but provides the least basis for generalization.

Snowball sampling asks initial participants to refer others who meet the study criteria. It is particularly useful for accessing hidden or stigmatized populations. However, it tends to produce samples that cluster around social networks, potentially missing isolated individuals.

Quota sampling identifies key subgroups and recruits predetermined numbers from each, resembling stratified sampling but without random selection within strata. It ensures representation of key groups but shares the generalizability limitations of other non probability methods.

Because selection probabilities are unknown, any estimate for a wider population from a non probability sample rests on a statistical model and its assumptions, for example that weighting or quotas on known characteristics remove the differences between sample and population that matter for the outcome. The American Association for Public Opinion Research (AAPOR) task force on non probability sampling concluded in 2013 that such samples may support statistical inference, but that the validity of the inference rests on how appropriate those assumptions are, and that no single framework covers all non probability methods.

Use non probability sampling when the research goal is to understand a phenomenon in depth rather than to estimate population parameters, when the population is difficult to enumerate or access, when the study is exploratory or pilot in nature, or when time and budget constraints make probability sampling impractical. Purposive sampling is standard in qualitative research. Convenience sampling may be acceptable in early stage quantitative research if its limitations are acknowledged. Snowball sampling is often the only feasible approach for studying hidden populations. When generalizability to a defined population is the goal, probability sampling should be used whenever possible.

A serious mistake is using a non probability sample and then making claims about population parameters (such as prevalence or mean values) without acknowledging the limitations of the sampling method. Another error is assuming that a large non probability sample compensates for the lack of random selection; a large convenience sample does not become representative simply because it is big. Researchers sometimes fail to describe and justify their sampling method, making it impossible for readers to evaluate the potential for selection bias. In qualitative research, a common mistake is using convenience sampling when purposive sampling would yield more informative data. Finally, snowball samples that rely on a single initial contact can produce heavily networked samples that miss isolated or marginalized individuals.

A qualitative researcher studying the experiences of refugees resettling in a mid sized American city uses purposive maximum variation sampling to recruit 20 participants from five different countries of origin, three age groups, and both genders. Participants with these characteristics are identified through community organizations, faith based groups, and social service agencies. The researcher also uses snowball sampling to reach three additional participants from an underrepresented country of origin. The sampling strategy is documented in the methods section, including the criteria for selection, the rationale for seeking maximum variation, and the final composition of the sample.