Individual Change Models for the Analysis of Randomized Longitudinal Designs: The Role of the Pretest

• Activity Number: 415
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Joseph R. Rausch*+
• Companies: Cincinnati Children's Hospital Medical Center
• Keywords: randomized ; longitudinal ; pretest ; individual change ; clinical trial ; multilevel models

Abstract

Randomized longitudinal designs provide a number of advantages over randomized pre-post designs, including the ability to employ individual change models for the examination of treatment effects. In this context, individual change models allow for a more precise specification of the treatment effect and increased power and precision. However, the appropriate role of the pretest within such models has generally been unclear in the literature. The present talk provides clarity on this issue, including how the models of interest can be fit within a multilevel modeling framework. Furthermore, the models of interest are compared with respect to statistical validity and efficiency in complete and missing data scenarios. It is concluded that researchers should generally use (a) the pretest as part of the outcome vector and (b) some form of the pretest as a covariate when using individual change models to examine treatment effects in randomized longitudinal designs.