When devising a course of treatment for a patient, doctors often have little\nquantitative evidence on which to base their decisions, beyond their medical\neducation and published clinical trials. Stanford Health Care alone has\nmillions of electronic medical records (EMRs) that are only just recently being\nleveraged to inform better treatment recommendations. These data present a\nunique challenge because they are high-dimensional and observational. Our goal\nis to make personalized treatment recommendations based on the outcomes for\npast patients similar to a new patient. We propose and analyze three methods\nfor estimating heterogeneous treatment effects using observational data. Our\nmethods perform well in simulations using a wide variety of treatment effect\nfunctions, and we present results of applying the two most promising methods to\ndata from The SPRINT Data Analysis Challenge, from a large randomized trial of\na treatment for high blood pressure.\n