Which of the following types of sampling is likely to have least bias?

Concept. Sampling designs fall into two families. In probability sampling every unit of the population has a known, non-zero chance of being selected and the selection is carried out by a chance mechanism; in non-probability sampling the units that enter the sample are decided by convenience, by the researcher's judgement, or by referrals from people already in the study. Systematic selection bias arises whenever the rule that admits a unit into the sample is itself related to the characteristics being measured, so bias is minimised only when that admission rule is independent of those characteristics — which is exactly what a chance mechanism guarantees.

Application. Simple random sampling draws units from a complete sampling frame using a lottery, a random-number table, or a random-number generator. Every unit carries the same probability of selection and each draw is independent of the researcher's preferences and of the respondents' social ties, so no subgroup is systematically over-represented or under-represented. What is left is random sampling error, which is unsystematic, measurable through the standard error, and shrinks as the sample size grows.

Contrast with the other designs.

• Snowball sampling reaches new participants only through referrals from earlier ones, so the sample travels along existing social networks and well-connected people are over-represented while isolated members may never be reached.

• Convenience sampling admits whoever happens to be available and easy to approach, so availability — usually itself related to the variables under study — decides membership.

• Purposive sampling admits the units the researcher judges to be typical or especially informative, so the researcher's prior expectations shape the composition of the sample.

Cross-check. Among the four designs listed, only simple random sampling is a probability design; in the other three the selection probability of a unit is unknown. Because a known, chance-governed selection probability is the defining safeguard against systematic selection bias, simple random sampling is the design likely to have the least bias here.