Parametric
Analysis of covariance (ANCOVA)
Compares group means adjusted for one or more covariates.
Variables
- Dependent variableContinuous — 1
- FactorNominal · Ordinal — 1
- CovariatesContinuous — at least 1
What it reports
Covariate-adjusted means; F, df, p; partial η² with a 90% CI.
statsmodels · Open the accuracy report
When does this apply?
For a design in which group means are compared after adjustment for one or more continuous covariates — comparing a final score while accounting for a baseline, say. At least one covariate is required: without one this is an ordinary one-way ANOVA. The headline result is the factor’s adjusted effect; the covariates are in the model to remove variance. The factor-by-covariate interaction is reported as a diagnostic from a separate model and never replaces the model.