A0743
Title: Robust functional-input kernel machine test for identifying a set of functional variables
Authors: Mengkun Chen - Virginia Polytechnic Institute and State University (United States) [presenting]
Inyoung Kim - Virginia Tech (United States)
Abstract: We propose a powerful and robust testing procedure for identifying significant functional variables associated with a continuous response variables related to certain disease or symptoms. Traditional methods were developed under parametric and nonparametric models assuming that functional variables are independent and the response variable follows parametric distribution such as normal distribution. The limitation of these methods may miss subtle changes at each functional variable by ignoring the dependence among functional variables. Also, they may be affected by outliers when the response variable follows heavy tailed or skewed distribution. To overcome this limitation, we propose to use semiparametric quantile regression to model the complex association and use permutation likelihood ratio test to identify the significant set of functional variables. We evaluate the performance of our method using simulation studies and apply the method to the ASD data set.
