Samplingthe spoonful and the soup
Measure a few, conclude about all — why a stirred spoonful beats drinking the pot, and what bias breaks.
Before this
End of this line
Pick a new destinationThis is the last stop in Statistics minThis opens up
Nothing yet. No lesson builds on this one so far.
See alsoTemperature
Transcript
It follows the voice. Click a sentence to jump there.
A cook wondering whether the soup needs more salt does not drink the whole pot. She gives it a good stir, lifts out one spoonful, and tastes that. If the pot is well mixed, whatever is true of the spoonful — too salty, too bland — is true of the whole batch.
That kitchen move has a name: sampling. Measure a few, conclude about all. The whole group you want to know about — every voter, every fish in the lake — is the population. The handful you actually measure is the sample. Ask a thousand people, chosen fairly, and their answers can stand in for a country of millions.
Two things make the spoonful trustworthy. First, the stir: every drop needs an equal chance of landing in the spoon, and for people that means picking names at random — random selection. Second, size. Somewhere sits the true answer. A random sample of ten gives a wide, wobbly band of guesses around it. A hundred narrows the band. A thousand squeezes it tight. Bigger samples simply wobble less.
One thing breaks it: bias. Skip the stir and scoop from one corner of the pot, where the salt has settled, and the spoonful stops copying the whole. Poll a single neighborhood and you hear one kind of voice, however many doors you knock on. The estimate lands in a tight band around the wrong value. A bigger sample from the same corner is just more confidently wrong — no size can fix an unfair scoop.
So a good sample is a small, honest copy of the whole — the same mix as the pot, just less of it. Size buys steadiness, but only fairness buys truth. The cook never had to drink the pot; one stirred spoonful told her everything, because the power was in the stir, not the spoon.