Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 497 - Variable Selection
Type: Contributed
Date/Time: Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #313213
Title: Feature Screening for High-Dimensional Quadratic Generalized Linear Model via Point Biserial Correlation
Author(s): Jinzhu Jiang* and Junfeng Shang
Companies: Bowling Green State University and Bowling Green State University
Keywords: point-biserial correlation; high-dimensional generalized linear model; variable selection

Sure independence screening (SIS) is a well-known procedure in linear model feature selection for high and ultrahigh dimensional data based on the Pearson correlation (Fan and Lv, 2008). Yet, it is mainly focused on linear model with continuous response variable. In this paper, we considered extending the SIS method to high-dimensional quadratic generalized linear model with binary response variable. We developed an SIS procedure based on the Point-Biserial correlation, with the ability to find the correlation between a continuous variable and a binary random variable. The Point-Biserial sure independence screening (PB – SIS) can be implemented in a straightforward way as the original SIS procedure, but it works more specifically on high-dimensional generalized linear model with interaction terms. We established the sure screening property for PB-SIS methods and conducted simulation studies to evaluate its performance. The simulation studies showed that PB-SIS performs better that some other traditional generalized linear model selection methods. We also demonstrated its use with the application to a real data example.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program