Survey sampling

Survey sampling selects individuals from a population to estimate characteristics of that population. Its design determines how well estimates represent the population and how their uncertainty can be measured.

Sampling frame: A list or other operational representation of the population from which a sample is selected. It defines which population members can actually enter the sample.
Statistical population: The complete set of units about which a statistical analysis seeks to draw conclusions. Sampling begins by specifying which units the survey aims to represent.
Nonprobability sampling: A sampling approach in which selection probabilities are unknown or not governed by random selection. It contrasts with designs that support direct, design-based inference from known selection chances.
Horvitz–Thompson estimator: An estimator of a population total formed by weighting each sampled value by the inverse of its inclusion probability. It shows how unequal selection chances can still yield an unbiased total estimate.
Simple random sampling: A design in which every possible sample of a fixed size has the same probability of selection. It provides the baseline probability design for estimating population characteristics.
Probability sampling: A sampling approach in which each population unit has a known, nonzero chance of selection. Known selection chances support defensible population estimates and sampling-error calculations.
Census: A complete enumeration that collects information from every unit in a defined population. It avoids sampling error but can still suffer from nonresponse and measurement error.
Margin of error: A stated range around a survey estimate that reflects a chosen confidence level and its sampling uncertainty. It communicates precision derived from the sample design and estimator.
Stratified sampling: A design that divides a population into groups and samples separately within each group. It can ensure coverage of important subgroups and improve precision.
Inclusion probability: The probability that a particular population unit is included in a sample. It underlies selection weights and design-based estimation.

Linked from 18 pages