|
Activity Number:
|
165
- SLDS CSpeed 2
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
Section on Statistical Learning and Data Science
|
|
Abstract #319089
|
|
|
Title:
|
Two-Stage Procedure for Efficient and Robust Inference on Heterogeneous Treatment Effect in Randomized Clinical Trials
|
|
Author(s):
|
Heng Chen* and Michael LeBlanc and James Dai
|
|
Companies:
|
Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center and Fred Hutchinson Cancer Research Center
|
|
Keywords:
|
Augmentation;
Function Optimization;
Nonparametric;
Robustness and Efficiency;
Permutation Test
|
|
Abstract:
|
Motivated by modified covariate approach, we propose a robust and efficient two-stage procedure for estimating heterogeneous treatment effects and testing for no heterogeneity in randomized clinical trials. Target estimands for HTE are defined in the mean difference scale for quantitative outcomes, and the risk ratio scale for binary outcomes. The first stage is to estimate the main-effect equivalence of baseline covariates on the outcome; the second stage is to estimate HTE with the first-stage main-effect estimator as an efficiency-augmentation term. This procedure allows adaptive nonparametric learning methods for both the main-effect component and HTE component. We prove that the proposed two-stage procedure is robust to model misspecification of main effects and achieve full efficiency with correct model specification A permutation test is proposed for high-dimensional HTE inference. As an illustrative example, we adopt gradient boosting trees (GBT) algorithm in the two-stage procedure for flexibly estimating heterogeneous treatment effects. Simulations and an application to a genetic study in the Prostate Cancer Prevention Trial are conducted to showcase these properties.
|
Authors who are presenting talks have a * after their name.
