Estimating Complier Average Causal Effect With Latent Growth Modeling

• In Event: 54.081 - Poster Session 11
  In Poster Session: 54.081-7 - Advances and Applications in Structural Equation Modeling

Abstract

When there is noncompliance during longitudinal experiments, we can estimate longitudinal Complier-Average Causal Effects (CACE) using the Standard Instrument Variable (IV) method or the Mixture Model Based (MMB) method under the Latent Growth Model (LGM) framework. However, it is unclear which of the two approaches is better and what factors might influence the accuracy and efficiency of the estimation. Therefore, the present study aims to answer three research questions through a simulation design: sample size, effect size, reliability of the observed variables, compliance composition, and outcome distribution, how will they each affect the Standard IV and the MMB estimators in term of their 1) accuracy, 2) statistical power, and 3) Type I error rate when effect size is 0?

Author

• Huili Liu, Insight Policy Research, Inc.