A0598
Title: A concave pairwise fusion approach to multiresponse regression clustering
Authors: Yuexiao Dong - Temple University (United States)
Abdul-Nasah Soale - Case Western Reserve University (United States) [presenting]
Chen Chen - Temple University (United States)
Abstract: Classical multiresponse regression studies the effect of a set of predictors on multiple responses. The problem is considered that the samples from the regression model consist of subgroups with different mean values and the same predictor effects. A concave penalized regression approach is used to detect the subgroups or clusters by shrinking the pairwise differences of the mean values. The proposed method exploits heterogeneity and automatically divides the observations into subgroups. The new method's performance is compared to the univariate response regression clustering method in simulation studies. An analysis of healthy and Parkinson's disease patients is also provided.