Indirect Multiple Response Regression

• Activity Number: 656
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Aaron Molstad* and Adam Rothman
• Companies: University of Minnesota and University of Minnesota
• Keywords: multiple response regression ; sparse covariance estimation ; abundant regression

Abstract

We propose a new class of estimators of the multiple response regression coefficient matrix that exploits the assumption that the responses and predictors have a joint multivariate normal distribution. These estimators do not require the popular assumption that the regression coefficient matrix is sparse or has small Frobenius norm. Using simulation studies, we show estimators from our class outperform relevant competitors under some data generating models. Two real data applications are presented and suggest that estimators in our class are competitive with existing methods.