Primary Submission Category: Randomized Studies
Identifying localized treatment effects in high-dimensional outcome spaces
Authors: Yujin Jeong, Ramesh Johari, Emily Fox,
Presenting Author: Yujin Jeong*
Based on technological advances in sensing modalities, randomized trials with primary outcomes represented as high-dimensional vectors have become increasingly prevalent. For example, these outcomes could be week-long time-series data from wearable devices or neuro-signal graph data derived from magnetic resonance imaging. This paper focuses on randomized treatment studies with such high-dimensional outcomes characterized by localized treatment effects, where interventions may influence a small number of dimensions, e.g., small temporal windows or specific clustered brain regions. Conventional practices, such as using fixed low-dimensional summaries for outcomes, result in significantly reduced power for detecting treatment effect signals. To address this limitation, we propose a procedure that involves subset selection followed by inference. Specifically, given a set of direction vectors, we identify the subset that captures treatment effects and subsequently conduct inference on these selected directions. Via theoretical analysis as well as simulations, we demonstrate that our method asymptotically selects the correct subset and increases statistical power.