Development of Joint Estimating Equation Approaches to Merging Clustered or Longitudinal Datasets from Multiple Biomedical Studies.

Abstract

Jointly analyzing multiple datasets arising from similar studies has drawn increasing attention in recent years. In this dissertation, we investigate three primary problems pertinent to merging clustered or longitudinal datasets from multiple biomedical studies. The first project concerns the development of a rigorous hypothesis testing procedure to assess the validity of data merging and a joint estimation approach to obtaining regression coefficient estimates when merging data is permitted. The proposed methods can account for different within-subject correlations and follow-up schedules in different longitudinal studies. The second project concerns the development of an effective statistical method that enables to merge multiple longitudinal datasets subject to various heterogeneous characteristics, such as different follow-up schedules and study-specific missing covariates (e.g. covariates observed in some studies but completely missing in other studies). The presence of study-specific missing covariates gives rise to a great challenge in data merging and analysis, where methods of imputation and inverse probability weighting are not directly applicable. We propose a joint estimating function approach to addressing this key challenge, in which a novel nonparametric estimating function constructed via splines-based sieve approximation is utilized to bridge estimating equations from studies with missing covariates to those with fully observed covariates. Under mild regularity conditions, we show that the proposed estimator is consistent and asymptotically normal. The third project is devoted to the development of a screening procedure for parameter homogeneity, which is the key feature to reduce model complexity in the process of data merging. We consider the longitudinal marginal model for merged studies, in which the classical hypothesis testing approach to evaluating all possible subsets of common regression parameters can be combinatorially complex and computationally prohibitive. We develop a regularization method that can overcome this difficulty by applying the idea of adaptive fused lasso in that restrictions are imposed on differences of pairs of parameters between studies. The selection procedure will automatically detect common parameters across all or subsets of studies.