Matching is a process used mainly to gain efficiency in experimental design. For example, in a case control study, participants who experienced a disease of interest may be matched to healthy cases and a logistic regression model may be used to assess the relative risk for a particular exposure, like smoking.

Matching can reduce the sample size drastically. However, comparisons are more highly balanced leading to greater precision in carefully matched samples. Matching, like stratification or adjustment in regression modeling, allows a researcher to control for variables that may have confounding or blocking impact. Unlike adjustment, matching may relieve the need to make assumptions about the functional form of the matching factor and the outcome.

Matching may be specified 1-to-1 or 10-to-1 or any fixed proportion that the researcher feels to be appropriate. Specialized methods for analyzing matched data include a paired t-test, linear regression models with an offset, conditional logistic regression, McNemar's test, and heirarchical linear models.