Online Program

Dual-Centered ANCOVA: Modifying ANCOVA to Model Within-Person Change (309601)

Keywords: Dual-center ANCOVA, ANCOVA, difference scores, physical health

A primary rationale for statistical controls in longitudinal research is to remove selection bias from regression coefficients relevant for making causal inferences in nonrandomized studies. Difference scores and residual change scores are the two fundamental approaches in analyzing change in longitudinal studies but can produce contradictory results in nonrandomized studies, illustrated by Lord’s paradox. The inconsistency between the two types of change methods is due to a violation of one of the assumptions of ANCOVA, namely independence of covariate and treatment. The current study introduces a novel dual-centered ANCOVA to remove the pretest difference between the treatment and control groups. I used longitudinal data on medical treatments and subsequent physical health to show the inconsistencies between the two types of change analyses and to show how to use dual-centered ANCOVA to remove the pretest difference to model within-person change and thereby to achieve consistent results between the two types of change analyses.