Random Samples and Which Population a Result Extends To

A random sample is a small stand-in for a bigger group, but only for the exact group it came from. If a question polls seniors at one school, the result speaks for seniors at that school, not for teenagers everywhere, because no student outside that school ever had a chance of being picked. The rule is simple: look at the list the sample was actually drawn from, called the sampling frame, and the conclusion extends exactly that far and no further. Draw from one gym's membership and you can speak for that gym. Draw from one city and you can speak for that city, not the whole state or country.

This trips people up because size feels like it should matter more than it does. A pile of ten thousand volunteer survey responses collected online looks more convincing than a random sample of three hundred, but volunteers choose themselves, and self-selection usually tracks with the opinion or behavior being measured. That bias doesn't shrink as more volunteers join; it just gets repeated at a larger scale. Random selection is what lets a sample substitute for a population at all. Sample size only controls how much the result might bounce around due to chance, which shows up later as margin of error.

This idea sets up margin of error and causal claims right after it. Once you know which population a sample supports, margin of error tells you how precise that estimate is, and whether a study was randomly sampled versus randomly assigned decides whether you can call the result a cause or just an association.