Guide
Sampling methods
Sampling methods are the techniques researchers use to select a subset of a population to study. They divide into probability methods, which use random selection, and non-probability methods, which do not.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Why sampling matters
Researchers rarely study an entire population — it is usually too large, costly, or impractical. Instead they select a sample, a manageable subset, and use it to draw conclusions about the whole. The quality of those conclusions depends heavily on how the sample was chosen. A well-chosen sample is representative: its composition reflects the population, so findings can be generalised back to it with quantifiable confidence. A poorly chosen one introduces sampling bias, where some members are systematically over- or under-represented, distorting results no matter how large the sample. The first decision is therefore which family of methods to use: probability or non-probability sampling.
Probability sampling
In probability sampling, every member of the population has a known, non-zero chance of being selected, achieved through random selection. This is what licenses statistical inference and the calculation of margins of error. Simple random sampling gives every unit an equal chance, like drawing names from a hat. Systematic sampling selects every nth unit from an ordered list after a random start — efficient, but risky if the list has a hidden periodic pattern. Stratified sampling divides the population into subgroups (strata) and samples within each, improving precision and guaranteeing representation. Cluster sampling divides the population into naturally occurring groups (clusters), randomly selects whole clusters, and surveys everyone (or a further sample) within them — cheaper for geographically dispersed populations but usually less precise.
Non-probability sampling
Non-probability sampling selects units by some criterion other than chance, so the probability of inclusion is unknown and formal generalisation to the population is not justified. Convenience sampling takes whoever is easiest to reach, such as students in a class — fast and cheap, but highly vulnerable to bias. Purposive (judgemental) sampling deliberately selects cases that fit the research aim, for example experts on a topic, and is common in qualitative work. Quota sampling fills preset quotas for subgroups (say, equal numbers of men and women) without random selection within them. Snowball sampling asks existing participants to recruit others, which is invaluable for hard-to-reach or hidden populations but tends to over-represent well-connected individuals.
How to choose a method
The right method follows from the research goal, the resources available, and the access you have to the population. If the aim is to produce estimates that generalise to a defined population with measurable uncertainty — a prevalence survey, an opinion poll — a probability method is needed, and a complete sampling frame (a list of the population) is required to implement it. If the aim is to explore a phenomenon in depth, study a specific group, or reach a population with no available frame, a non-probability method is appropriate and often unavoidable. Budget, time, and the cost of reaching dispersed units also weigh on the decision: cluster sampling trades some precision for lower cost, while stratified sampling buys precision at the cost of needing detailed frame information.
Sample size, representativeness and bias
A representative sample reflects the population’s relevant characteristics; representativeness comes chiefly from the selection method, not merely from size. A large convenience sample can still be badly biased, while a modest probability sample can be sound — which is why simply collecting more responses does not fix a flawed method. Sample size affects precision: larger probability samples narrow the margin of error and improve the power to detect real effects, with the right size determined by the desired precision, expected effect size and population variability rather than a fixed rule of thumb. Beware of bias at every stage: undercoverage (parts of the population missing from the frame), non-response (those who decline differing systematically from those who answer), and self-selection. Good practice — sound frames, randomisation where possible, and transparent reporting of how the sample was drawn — supports the generalisability and external validity of the findings.
Key facts
At a glance
- Definition: techniques for selecting a subset of a population to study
- Two families: probability (random selection) and non-probability (no random selection)
- Probability types: simple random, systematic, stratified, cluster
- Non-probability types: convenience, purposive, quota, snowball
- Generalisation: only probability sampling supports formal statistical inference
- Representativeness: driven mainly by selection method, not sample size alone
Common questions
FAQ
What is the difference between probability and non-probability sampling?+
In probability sampling, every member of the population has a known, non-zero chance of selection through random choice, which allows results to be generalised to the population with measurable uncertainty. In non-probability sampling, selection relies on convenience or judgement, the chance of inclusion is unknown, and formal generalisation is not justified — though such methods are quicker and suit exploratory or hard-to-reach populations.
Which sampling method is best?+
There is no single best method; the right choice depends on the research aim, resources and access to the population. If you need estimates that generalise to a defined population, use a probability method such as simple random or stratified sampling. If you are exploring a phenomenon, studying a specific group, or have no sampling frame, a non-probability method such as purposive or snowball sampling is appropriate.
Does a larger sample remove bias?+
No. Sample size improves precision but does not correct a biased selection method. A large convenience sample can be systematically unrepresentative, while a smaller probability sample can be sound. Reducing bias depends on how units are selected — using a good sampling frame and randomisation — and on minimising undercoverage, non-response and self-selection, not on collecting more responses.
