Description:

Department of Preventive Medicine
Feinberg School of Medicine
Northwestern University

“Interaction trees for Exploring Stratified and Individual Treatment Effects”

Assessing heterogeneous treatment effects has become a growing interest in many application fields. The concepts of stratified and individual treatment effects involved in such as assessment are closely related to subgroup analysis, personalized medicine, and optimal treatment regime. Concerning experimental data collected from randomized trials, we put forward a tree-structured method, termed “Interaction Trees” (IT), to explore stratified and individual treatment effects. IT recursively splits data in such a way that the treatment effect become more homogeneous within each resultant partition. The automated procedure facilitates a number of objectively defined subgroups, in some of which the treatment effect is found prominent while in others the treatment has a negligible or even negative effect. The standard CART (Breiman et al., 1984) methodology is inherited for built-invalidation in order to avoid false positive findings. Interference at different levels can be made on the basis of IT. An aggregated grouping procedure stratifies data into refined groups where the treatment effect remains homogeneous. Ensembles of IT models can provide prediction for individual treatment effects and it compares favorably to the traditional ‘separate regression methods. In order to extract meaningful interpretations, we also made available several other features such as variable importance and partial dependence plot. An empirical illustration of the proposed techniques is made via and analysis of quality of life (QOL) data among breast cancer survivors.

Xiaogang Su, PhD
Associate Professor
Department of Mathematical Sciences, University of Texas at El Paso

Monday, September 22, 2014
3 to 4:30 p.m.

Stamler Conference Room
Suite 1400, 14th Floor
680 North Lake Shore Drive
Chicago, Illinois 60611