Nonparametric and Semiparametric Approaches to Studying Longitudinal Patterns

• Activity Number: 368
• Type: Invited
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Naisyin Wang*+
• Companies: University of Michigan
• Keywords: longitudinal data analysis ; functional data analysis ; mixture modeling ; classification

Abstract

It is often of interest to study the similarity and differences of patterns of longitudinal observations from two populations, e.g. diseased versus normal ones. However, unlike the regular multivariate-variables, there are particular issues arise during the analysis of a longitudinal study. For example, the measurements may not be observed at the pre-determined and specific time points so that the longitudinal variables can be readily transformed to multivariate-variables. Furthermore, time scales might vary from subject to subject. In this study, we combine approaches from the traditional longitudinal and functional data analysis with data mining approaches to study the similarity and discrepancies of longitudinal patterns found in two different populations. The systematic investigations shed light on how the two populations differ in the global and individual levels as well as how they share similar structures. Numerical advantages of the proposed methods would be illustrated through simulation and several field data examples.