A comparison of methods for longitudinal analysis with missing data

Abstract

In a longitudinal two-group randomized trials design, also referred to as randomized parallel-groups design or split-plot repeated measures design, the important hypothesis of interest is whether there are differential rates of change over time, that is, whether there is a group by time interaction. Several analytic methods have been presented in the literature for testing this important hypothesis when data are incomplete. We studied these methods for the case in which the missing data pattern is non-monotone. In agreement with earlier work on monotone missing data patterns, our results on bias, sampling variability, Type I error and power support the use of a procedure due to Overall, Ahn, Shivakumar, and Kalburgi (1999) that can easily be implemented with SAS's PROC MIXED. Copyright © 2004 JMASM, Inc.