Unified Modeling Methods for Curve Predictors and Different Responses

• Activity Number: 674
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Xiaohui Wang*+ and Veera Baladandayuthapani and Bani K. Mallick and Kim-Ahn Do and Shubhankar Ray
• Companies: The University of Texas-Pan American and MD Anderson Cancer Center and Texas A&M University and MD Anderson Cancer Center and Merck & Co., Inc.
• Keywords: Unified modeling ; Bayesian methods ; Curve predictors ; Classification ; Survival response

Abstract

A lot of biological studies involve curve data that are functions in nature and used as predictor. The responses of interest are either classification labels or survival outcomes. Analysis of curve predictors with those responses can be very challenging because of the features of different data. We studied both cases with unified Bayesian modeling methods. In our models, we used wavelet- and spline-based nonparametric approaches wherein the usage of those bases functions simplifies the parameterizations and the unified modeling framework allows synergistic benefit between the regression of curve predictors and modeling of response data. Examples of application on real world data sets, such as a prostate cancer clinical trial study and a study on rats for detection of doxorubincin-induced cardiotoxicity based on serum spectra curves, are illustrated.