Workshop 10: Testing for and characterizing treatment effect heterogeneity under the Neyman-Rubin potential outcomes framework

The 10th 2013/2014 Workshop on Quantitative Methods in Education, Health and the Social Sciences (QMeHSS) will be on Friday May 30th from 11:00-12:30 and will be led by Dr. Luke Miratrix. The seminar will be held in the NORC conference room 344. NORC is located at 1155 E. 60th Street.

 

Testing for and characterizing treatment effect heterogeneity under the Neyman-Rubin potential outcomes framework

Luke Miratrix, Ph.D.

Department of Statistics, Harvard University

Many forms of analysis of experimental data are either implicitly or explicitly build on an assumption of constant treatment effect.  Other approaches estimate average effect but are agnostic as to whether that effect varies by unit.  While such omnibus assessments of treatment are important, understanding treatment effect variation, i.e., how a treatment differentially impacts subjects, is also important because it allows for a richer understanding of an overall causal effect of some intervention.  This is particularly relevant for personalized medicine, where we wish to only treat those who would benefit from a drug, or social service interventions, where there are finite resources which limit the number of units that could potentially be treated and we wish to maximize overall impact. We design and explore some nonparametric tests inspired by permutation and resampling methods in conjunction with the Neyman-Rubin potential outcomes framework to test for treatment effect heterogeneity.  In particular, we focus on using the Kolmogorov-Smirnoff (KS) statistic in conjunction with different methods for estimating the average mean effect, a nuisance parameter in our context.  These tests can then be extended to assess heterogeneity beyond any effects explained by covariates via, in the discrete case, individual treatment effect estimates by strata or, in the continuous case, a regression model.