Testing a High-Dimensional Parameter in the Presence of High-Dimensional Nuisance Parameters

• Activity Number: 397 - Personalized Medicine with Large-Scale Data: Beyond Machine Learning
• Type: Invited
• Date/Time: Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Wei Pan*
• Companies: University of Minnesota
• Keywords: Adaptive tests; Deep learning; Machine learning; Statistical significance; Truncated Lasso penalty (TLP)

Abstract

A key to personalized medicine is to detect interactions between a treatment and individual-specific characteristics, the latter of which is represented by a high-dimensional vector of demographic, clinical and genetic information (with induced high-dimensional nuisance parameters). Although there are some machine learning algorithms for this purpose, often they cannot be used for statistical inference, e.g. statistical significance testing, on such detected interactions. We develop such a test and show its performance and application.