Online Program Home
My Program

Abstract Details

Activity Number: 529 - Regression Trees and Random Forests
Type: Contributed
Date/Time: Wednesday, August 1, 2018 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistical Learning and Data Science
Abstract #329851
Title: Conditional Quantile Regression Tree/Random Forest
Author(s): Huichen Zhu* and Ying Wei
Companies: and Columbia University
Keywords: Quantile Regression; Regression Tree; Random Forest; Precision Medicine

Classification and Regression tree is a classical statistical learning method, which recursively partitions the sample space into mutually exclusive sub-spaces with means of an outcome of interest. Simple visualization and ability dealing with a large number of potential predictors of regression tree are important considerations in the application to precision health. Existing mean-based partitioning algorithm may not be effective to identify the most vulnerable subpopulation. We propose a conditional quantile-based tree/random forests, which provides a systematic strategy for examining how covariates influence various quantile regions of the outcome distribution. The asymptotic property of the proposed method and numeric study indicates that it gives a good estimation of treatment effect. We apply our proposed method the Reactions to Acute Care and Hospitalization data, which includes patients with Post-Traumatic Stress Disorder and other emergency room related experience variable. The length of prediction interval and coverage percentage show that the proposed method achieves to get accurate predictions for those patients with severe PTSD.

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

Back to the full JSM 2018 program