An Evaluation of Growth Models With Propensity Score Weighting for Longitudinal Observational Data

• In Event: Roundtable Session 15
  In Roundtable Session: Nuanced Methods for Latent Growth Models

Abstract

This study evaluates latent growth models (LGM)combined with three methods to obtain propensity score weights (odds of treatment, one-to-many matching with replacement within calipers, and optimal full matching) to estimate treatment effects in longitudinal observational studies. An applied example is provided where the effect of obtaining a GED on earnings is estimated using data from the National Longitudinal Survey of Youth (NLSY97). A simulation study examines the performance of LGM with propensity score weights with respect to bias of treatment effect estimates and standard errors.

Authors

• Walter L. Leite, University of Florida

• Sungur Gurel, University of Florida

• Laura M. Stapleton, University of Maryland