Studying an entire population is often impractical. Sampling selects a subset of individuals so that conclusions can be drawn about the whole. The quality of those conclusions depends on how the sample is chosen. Probability samples, in which every member has a known chance of selection, support statistical inference. Non-probability samples are easier to obtain but limit generalizability.

Sample size affects precision: larger samples generally produce narrower confidence intervals. Diminishing returns set in, so the optimal size balances cost against the needed accuracy. Stratification, clustering, and other design features can improve efficiency or practicality while preserving the ability to make valid inferences.

Sampling is fundamental to survey research, quality control, and many scientific studies. Understanding its principles helps both producers and consumers of research judge how much weight to give the findings. A well-designed sample of manageable size can be more informative than a large but biased collection of observations.