An MVA Approach to Missing $E_T$

We present an overview of an analysis of the missing transverse energy (MET) resolution in the CMS detector studying $Z\to\mu\mu$ decay events as a part of the University of Michigan Summer REU Program. We make use of boosted decision trees and specialized recoil variables in an attempt to isolate separate effects contributing to the detector resolution such as pileup, the underlying event, and detector inefficiencies. A metric is developed to study the body of high-MET events in the distribution. Emphasis is placed on constructing a multivariate analysis that effectively models and corrects these `tail' regions, resulting in an overall improvement in our MET resolution. In Monte-Carlo simulation, we see an improvement from a mean value of MET = 18.12 GeV to MET= 8.14 GeV with full MVA corrections.