Targeting Toward Inferential Goals for Enhancing Subject-Specific Growth Parameter Estimation in Longitudinal Data Analysis

• In Event: Application of Longitudinal Models and Statistical Theory

Abstract

In longitudinal studies, targets of inference often include obtaining individual subject-specific growth parameters, determining their rankings, and examining their distribution across multiple subjects. This paper explores strategies to enhance inferences on subject-specific growth parameters in longitudinal studies, including flexible semiparametric modeling to relax normality assumption, and posterior summarization methods tailored to specific inferential goals. Our simulation reveals a trade-off: estimators optimal for individual subject-specific growth may yield subpar estimates for growth parameter distribution, and vice versa, contingent on data reliability levels. This underscores the need to carefully balance trade-offs associated with shrinkage when designing and analyzing longitudinal studies.

Author

• Joon-Ho Lee, University of Alabama