Which of the following sampling methods is not based on probability ?

Concept: Sampling designs fall into two families according to how units actually reach the sample. In probability sampling a chance mechanism does the selecting, so every unit of the population carries a known, non-zero chance of inclusion and the sampling error can be estimated. In non-probability sampling the investigator or field-worker does the selecting on grounds of convenience, judgement or availability, so an individual unit’s chance of inclusion is unknown and the sample cannot be projected to the population with a measurable margin of error.

Application: Apply one criterion to each design offered — at the moment an individual unit enters the sample, is that choice made by a chance mechanism or by a person?

• Simple random sampling — the population is treated as one undivided frame and units are drawn from it by lottery or random-number tables; the final choice of unit is made by chance.

• Stratified sampling — the frame is first divided into internally homogeneous strata, and an independent random draw is then made inside each stratum; the final choice of unit is made by chance.

• Cluster sampling — the frame is divided into naturally occurring groups and whole groups are drawn at random, every unit of a selected group entering the sample; the choice of group is made by chance.

• Quota sampling — the number of respondents required from each category is fixed in advance, but which particular persons fill those quotas is settled by the field-worker on grounds of convenience or judgement; the final choice of unit is made by a person, not by chance.

Cross-check: Quota sampling is commonly described as the non-probability counterpart of stratified sampling, because both partition the population into subgroups and fix how many units come from each. The two part company at the last step: stratified sampling draws at random within a stratum, whereas quota sampling leaves that selection to the investigator. Since the probability that any given person enters a quota is unknown, standard error and confidence intervals cannot be computed for a quota sample — the practical hallmark of a non-probability design.

Result: Among the designs offered, quota sampling is the one that is not based on probability.