valintabias

Valintabias, or selection bias, is a systematic distortion in estimates that arises when the subset of units selected for analysis is not representative of the target population due to nonrandom selection processes. It can affect the validity of conclusions in epidemiology, social science, and data analysis.

Common sources include nonrandom sampling, differential nonresponse, and self-selection, where the probability of inclusion relates to the outcome or exposure. Other forms include survival bias, where only individuals who survive to be observed are included; ascertainment bias, where measurement depends on group; and sampling frame errors.

For example, a survey that only reaches households with landline phones may underrepresent younger or lower-income groups; a clinical trial that enrolls volunteers may overrepresent health-conscious individuals.

Mitigation strategies include random sampling, stratified sampling, and using weights to adjust for known differences; design choices such as cohort or case-control studies; statistical methods like propensity score weighting, inverse probability weighting, and multiple imputation for missing data; and sensitivity analyses to assess potential bias.

Valintabias is often discussed alongside other biases such as measurement bias and confounding. In English-language literature selection bias is the common term; in Finnish contexts valintabias may be used to describe the same concept, highlighting the influence of sampling on study results.