Poststratified

Poststratified refers to data or samples that have undergone poststratification, a weighting adjustment used in survey sampling to align the sample with known population totals for one or more stratification variables, such as age, sex, region, or education. The aim is to reduce bias from differential response rates and sampling errors.

In practice, poststratification involves dividing the population into poststrata by cross-classifying selected variables. After collecting responses, the analyst compares the weighted counts in each cell to known population totals and adjusts weights accordingly. A common approach is to multiply each respondent’s current weight by the ratio of the population total to the sample total within their cell. Weights may then be normalized to preserve the overall sample size or total weight, and poststratification can be combined with other weighting methods like calibration or raking.

Applications are widespread in political polling, market research, and social surveys where reliable population benchmarks exist. Poststratification helps improve representativeness when nonresponse or coverage errors differ across subgroups, provided the poststrata capture relevant sources of variation.

Limitations include the need for accurate and stable population totals for the chosen variables; sparse cells can increase variance and reduce precision; and unobserved characteristics correlated with the outcome may still bias estimates. If the sampling frame is biased, poststratification cannot fully correct all issues.

Related concepts include calibration weighting, iterative proportional fitting (raking), and model-based poststratification, such as multilevel poststratification used in Bayesian analyses to improve small-area estimates.