13 Research methods, statistics, critical review and evidence-based practice13.7 Qualitative research
13.7.2 Qualitative sampling
3 min read1 figure
Written by Resident Doctor Committee · Published
Qualitative sampling
Purposive sampling
Purposive sampling selects participants deliberately, by judgement or nomination, because they are information-rich and have characteristics relevant to the research question.
It aims for relevant experience and a range of perspectives, not statistical representativeness.
Variant | Basis of selection |
|---|---|
Maximum variation | Participants span the widest range of outcomes and circumstances, and themes that hold across them are taken as central |
Intensity | A few information-rich informants who have lived the experience in depth but are not deviant cases, also called the key-informant approach |
Criterion | A specified experience, such as detention under section 3 |
Typical case | Cases judged ordinary, chosen to describe usual experience |
Theoretical | The emerging analysis decides who is recruited next, to develop categories in grounded theory |
Convenience sampling
Convenience (opportunistic) sampling recruits whoever is accessible and willing, first come first served, until a target number is reached, such as clinic attenders.
Accessibility then shapes the findings, because the sample misses non-attenders and people who are unavailable in working hours.
Applying basic eligibility criteria to whoever turns up does not make a convenience sample purposive.
Snowball sampling
Snowball sampling recruits through nominations and introductions from people already in the study.
It suits hidden, stigmatised or hard-to-reach groups for whom there is no sampling frame.
It over-represents well-connected people and misses isolated ones, so researchers use several starting points to broaden coverage.
Sample size
No universal number exists, and a study that claims saturation must define it and show it.
Malterud, Siersma and Guassora's information power framework says a study needs fewer participants when the aim is narrow, the experience fits the question closely, established theory supports it, the dialogue is rich and the analysis is deep rather than comparative.
Reporting the sample
Researchers report the recruitment source, the sampling strategy and its rationale, exclusions and refusals, so readers can judge transferability to other settings.
Data saturation
Saturation and sample size
Data saturation sets sample size in qualitative research. Researchers analyse data as they collect it and stop recruiting when new data yield no new themes.
A power calculation fixes sample size only in quantitative designs.
A study that fixed its sample size in advance cannot claim that saturation decided when recruitment stopped.
Three senses of saturation
Sense | What is exhausted | Where it belongs |
|---|---|---|
Code saturation | New relevant codes | Some codebook-based approaches |
Meaning saturation | Further depth about the variation and context of issues | Approaches that use the concept explicitly |
Theoretical saturation | Further development of categories | Grounded theory with theoretical sampling |
Theoretical saturation in grounded theory is reached when further data and constant comparison produce no new categories or properties.
Theoretical sampling recruits people who can develop categories that are still thin, rather than simply the next available participant.
In Hennink, Kaiser and Marconi's 25 interviews, code saturation came at about nine, but meaning saturation needed 16 to 24.
Limits
Guest, Bunce and Johnson found saturation within 12 of 60 interviews in one study, but 12 is not a general rule.
Apparent saturation may be false when interviews are thin or recruitment is narrow, for example clinic attenders only.
Braun and Clarke argue that saturation does not suit reflexive thematic analysis, where researchers generate themes rather than discover them.
Information power is an alternative rationale: the more relevant information a sample holds for a specific aim, the fewer participants it needs.