Inference on Mean Treatment Effects After Model Selection

• Activity Number: 552
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Jingshen Wang*
• Companies: University of Michigan
• Keywords: model selection ; treatment effect ; high dimensional data

Abstract

We discuss statistical inference on treatment effects after model selection. Following the idea of Efron (2014), we use bagging, also known as bootstrap smoothing, to tame the erratic discontinuities of selection-based estimators. A simple formula for the accuracy of bagging provides variance estimation for the smoothed estimator. We show that this approximation enables us to construct reliable confidence intervals on the mean treatment effect even when the ratio of the covariate dimension to the sample size is close to one.