Watch on YouTube

10:14

Does the sample stand for the population it came from?

Generalizability asks whether results from your sample stand for the population that sample came from. How you answer depends on what carries the claim. A probability sample carries it by design, through known selection probabilities. A convenience sample carries it only through a model's assumptions, which you must state. A study honest about belonging to one village carries nothing further, and that is a legitimate choice. For a causal effect, reach also depends on effect modification: whether the effect itself differs across people.

Also covered: probability sampling · poststratification · convenience samples · effect modification