An autoregressive latent change score model for randomized pretest, posttest, follow-up designs

Randomized pretest, posttest, follow-up designs test treatment effects, but popular approaches for analyzing them such as the ANOVA or ANCOVA models have limitations, including lower power for the former and the use of residualized change scores for the latter. Latent change score models (LCSMs) have been proposed to address these issues, but none addresses both simultaneously. We develop an autoregressive LCSM that preserves the change score interpretation without resorting to residualization and show that it performs at least as well if not better than competing alternatives using a small simulation study. We also illustrate how it compares to other approaches using data from children at high risk for autism spectrum disorders.