Sampling

Sampling means drawing a subset, called the sample, from within a larger set called the population. The goal is either to study the population through a small part of it, or simply to extract a few items chosen at random. The term therefore covers both the idea of representation and that of selection.

The most telling mental image is that of a large urn filled with balls: rather than examining them all, you remove a handful. The decisive question is then how you remove those balls, because the way you draw them determines the reliability of what you observe.

Two main modes exist. With replacement, each item drawn is put back before the next one, so the chances stay the same from one draw to the next and the same item can come up again. Without replacement, the item drawn stays out, the set shrinks, and the probabilities change with each removal. Drawing five cards from a deck of fifty-two falls under the second case: for the first card the chance of a given card is one in fifty-two, for the second one in fifty-one, and so on.

A common trap is to confuse a large sample with a good sample. Size does not correct a flaw in method: if the way of drawing favors certain items, the sample remains biased even when it is large. Only a truly random draw, where each item keeps an equal chance, gives a faithful picture of the population.

On the site, this principle is at work whenever a subgroup is drawn at random: selecting a few winners among participants, extracting a handful of names from a long list, or assembling a panel all rely on a sampling that you want to be free of bias.