1 General overview
Stratification involves, for the purpose of the statistical analysis, dividing participants into subgroups with each subgroup having the same or similar baseline characteristics. Stratied analyses thus enable like-for-like comparisons within the subgroups groups. e method is best used when there are only one or two baseline characteristics for which to stratify.2 Regression models estimate how each prognostic factor relates to the outcome (see Chapter 1.4). Linear regression is used for continuous outcomes, logistic regression for binary outcomes, and Cox models if censoring occurs. Propensity scoring models how these same covariates relate to treatment allocations rather than, as in regression models, outcomes. e underlying principle is that the propensity score summarizes the manner in which the baseline characteristics are associated with treatment allocations, so that selection bias can be removed when comparisons are made.2 Further details of these methods can be found in standard statistical texts.