A Comparison of Multivariate Methods for Measuring Change from Pre-Test to Post-Test

• Activity Number: 522
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Justin Rogers*+ and Daniel Mundfrom
• Companies: Abbott Laboratories and Eastern Kentucky University
• Keywords: Multivariate ; Change Scores ; Power ; MANOVA/MANCOVA ; Monte Carlo Simulation ; Pretest-Posttest Designs

Abstract

Three multivariate methods for measuring change from pretest to posttest are compared with respect to statistical power over various levels and combinations of effect size, alpha, sample size, number of dependent variables, number of significantly different dependent variables, correlation between corresponding pretest and posttest scores, and correlation between unrelated pretest and posttest scores. The MANCOVA method utilizing the posttest scores as dependent variables and pretest scores as covariates was found to have superior power in the majority of the scenarios examined. However, there were scenarios where the method utilizing change scores as the dependent variables and the method utilizing only posttest scores as the dependent variables displayed greater power. Using the results from the Monte Carlo simulations, comparisons are presented that reveal the conditions under which each of the three multivariate methods displayed greater statistical power than the other two. In addition to the immediate implications of the current study, suggested future avenues of research that could expand upon the current findings are discussed.