Abstract:
|
In high dimensional genomics studies, network—based penalization methods have achieved success since the performance of variable selection can be significantly improved with interconnections among genomic features being incorporated as network structures. Extensive network—constrained variable selection methods have been developed to accommodate a variety of outcome variables, such as the continuous disease phenotypes, discrete disease status and patients’ survival. However, its development in longitudinal studies is rather limited. In this study, we develop a novel network based variable selection method for longitudinal studies with high dimensional features. Extensive simulation studies have shown the superior performance of the proposed method over multiple alternatives. Analysis of a longitudinal study with SNP measurements has revealed its competitive practical performance.
|