A Regression Tree Approach to Identifying Subgroups with Differential Treatment Effects

Abstract

For hard-to-treat diseases, it is often difficult to discover new treatments that benefit all subjects.? A more realistic goal is to identify subgroups of subjects for whom the treatment has a large effect.? Regression trees are natural solutions because they partition the data space.? We propose two new methods that do not have selection bias and are applicable to data with censored responses, missing predictor values, and treatments with two or more levels.? Importance scores for identifying influential variables are obtained as well.? A bootstrap technique is used to construct confidence intervals for the treatment effects in each subgroup.? The new methods are compared with several existing methods in terms of accuracy and computational speed.?