ATLASResearch
methods
Quantitative/ Analysis

ANCOVA

Compare outcomes after defensible covariate adjustment

ANCOVA is a linear model combining group indicators with continuous covariates. In a randomised pre-test and post-test study, modelling follow-up with baseline adjustment can improve precision. Covariates require substantive justification and suitable functional forms. The traditional common-slope version assumes no group-by-covariate interaction, which should not be imposed blindly.

WHEN IT FITS

Use for continuous follow-up outcomes with informative baseline measurements or justified pre-exposure covariates. Specify adjustment before analysis and avoid conditioning on variables produced by the treatment.

Strengths

  • Can improve precision by using baseline information
  • Expresses adjusted differences in outcome units

Limitations

  • Post-treatment adjustment can introduce bias
  • Common slopes and extrapolated adjusted means can mislead

Know the boundary

ANCOVA adjustment does not make an observational comparison equivalent to randomisation.

USED ACROSS
Banking & financeBusiness & MBAPsychologyComputer scienceEducation