Convenience samples do not support unqualified population generalization
Aliases: convenience sampling · external validity · nonprobability sample
What it is
A convenience sample consists of people who are easy to reach and willing to participate: colleagues, students, current customers, or crowdworkers. It can yield valid observations about those participants and support pilots, problem discovery, and some controlled mechanism tests. Without additional justification, it cannot turn sample proportions, means, or preferences into numerical claims about a target user population.
Why it happens
Probability-based generalization needs known or defensibly modeled inclusion chances. Convenience inclusion is unknown and jointly shaped by geography, timing, relationships, devices, motivation, and research burden. Even a large sample has standard errors that describe random variation within the obtained data, not systematic differences created by selection. More participants cannot convert unknown coverage into representativeness.
Studying it
State the target population and the population actually reachable, then compare sample and benchmark distributions on variables related to the outcome. Quotas, propensity adjustment, and post-stratification require assumptions such as ignorable selection conditional on measured variables; report those assumptions and examine weight trimming and unmeasured selection sensitivity. Without a frame or credible adjustment variables, restrict claims to sample descriptions, associations, or mechanism tests. Replication across channels probes robustness without automatically restoring probability representation.
Where it stops holding
This limit does not make convenience samples worthless or require probability sampling for every inference. Randomized experiments can identify an effect within a convenience sample while leaving effect transport to other people unresolved. If the target population is exactly one course or one crowd platform, an accessible roster may nearly cover it. Statistical generalization requires sampling support; analytical or theoretical generalization instead rests on mechanisms, boundary conditions, and cross-case reasoning.