Variable Selection for Optimal Decision Making

This paper discusses variable selection for medical decision making; in particular decisions regarding when to provide treatment and which treatment to provide. Current variable selection techniques were developed for use in a supervised learning setting where the goal is optimal prediction of treatment response. These techniques often leave behind small but important interaction variables that are critical when the ultimate goal is optimal decision making rather than optimal prediction. While prediction of treatment response represents a first step in finding optimal decisions, this paper points out some key differences between prediction and decision making applications. The paper presents two new techniques designed specifically to find variables that aid in decision making and demonstrates the utility of these techniques on both simulated data and on real world data from a randomized controlled trial for the treatment of depression.