Which sampling type cannot provide inferences back to the larger population, no matter how large the sample?

• A. Probability
• B. Purposive
• C. Nonprobability
• D. Convenience

Answer: (C)

Generalizability hinges on whether the sampling method assigns known chances of selection to each member of the population. Only probability sampling provides a way to quantify representativeness and estimate sampling error, because every member has a known, nonzero chance of being chosen. This makes it possible to infer something about the population and attach confidence or margins of error to estimates.

Nonprobability sampling, by contrast, does not use random selection or known selection probabilities. Choices about who is included are driven by the researcher’s purpose or practical factors, not by randomization. Because there’s no way to assess how representative the sample is relative to the whole population, you can’t reliably generalize findings to that population—even with very large samples. The lack of a known probability model means you can’t quantify bias or error in the population estimate.

Within nonprobability approaches, purposive and convenience sampling fall into this category: they can be useful for exploring specific groups or contexts, but they don’t support broad population inferences in a defensible way. Probability sampling, such as simple random, stratified, or cluster designs, is what enables generalization with quantified uncertainty.