What is Sample Size?

Sample size is how many people or things a study actually measured. Small samples produce results that swing wildly by luck alone, which is why so many exciting small findings vanish when someone repeats them. But a large sample fixes only randomness, and it never fixes a badly chosen sample.

Every study measures a slice of the world and hopes that slice represents the whole, and the smaller the slice, the more luck decides what it sees. Toss a fair coin ten times and seven heads is ordinary; toss it a thousand times and seven hundred heads is extraordinary. Studies behave the same way, so a striking result from twenty participants is exactly what chance produces surprisingly often, which is one reason dramatic small-study findings vanish when other laboratories repeat the experiment. There is no universal right number, since it depends on the size of the effect and how much people differ. A national opinion poll needs around a thousand people for a margin of error near three points, whether the country holds five million people or three hundred million. Precision also gets expensive quickly, since halving that margin of error takes roughly four times as many participants. What size cannot buy is representativeness, so a sample that misses part of the population, such as surveying only young adults to understand all voters, stays biased however many you add. A million poorly chosen participants still produce a precisely wrong answer.

Judging a restaurant by one dish or by ten. One brilliant plate might mean the kitchen is superb, or might mean you got lucky with the chef's favourite. Ten plates across the menu tell you far more. But if all ten are desserts you still know nothing about the mains, because the problem was never how many dishes you ate: it was which ones you chose.