JSM 2012 Home

JSM 2012 Online Program

The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.

Online Program Home

Abstract Details

Activity Number: 613
Type: Contributed
Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
Sponsor: Section on Statistical Learning and Data Mining
Abstract - #304412
Title: Resampling-Based Test for Variable Selection with Gradient Boosting Trees
Author(s): Qianyi Zhang*+
Companies: Eli Lilly and Company
Address: Lilly USA, LLC, Indianapolis, IN, 46285, United States
Keywords: Variable Selection ; Resampling ; Permutation Testing ; Gradient Boosting ; Step-Down Multiple Comparison Procedure

Gradient boosting machine is a non-parametric ensemble model which aggregates many classification and regression trees. A valuable output of the method is estimated relative influences (RI) for all covariates. While RIs are helpful in identifying important covariates, formal significance testing with RIs is problematic. This paper proposes a resampling-based test to evaluate the significance of the RI. Probabilities of getting the RI as high as observed, when there is no relationship between the outcome and a given covariate are computed based on a reference null distribution. To address multiplicity inherent in evaluating a large number of candidate covariates, while taking into account the order of relative importance, a step-down multiple comparison procedure has been adopted. All covariates are ordered by their RI, from largest to smallest, and their associated multiplicity adjusted permutation based p-values are compared with a pre-specified alpha-level. Simulation studies were done to assess the properties of the proposed test and its robustness to the presence of highly correlated predictors.

The address information is for the authors that have a + after their name.
Authors who are presenting talks have a * after their name.

Back to the full JSM 2012 program

2012 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Continuing Education program, please contact the Education Department.