Study of Alternative Statistical Methods for a Model Unspecific Search in CMS

Since 2010, protons are collided with a center of mass energy of 7 TeV for the first time ever using the particle accelerator LHC. The Compact Muon Solenoid (CMS) is one of the experi- ments designed to measure the produced particles. Its purpose is to get a detailed picture of elementary particles in that regime. This includes the measurement of the parameters of the Standard Model and the search for new physics, not explained by the Standard Model. There are a number of models suggesting physics beyond the Standard Model. Many analy- ses of CMS data test whether a specific model can either be excluded by the data or if it explains the data more convincingly than the Standard Model. The Model Unspecific Search in CMS (MUSiC) analyses CMS data without considering such a model of new physics, but instead by comparing the data solely with predictions made by Standard Model Monte Carlo generators. In contrast to many conventional analyses MUSiC not only examines a small selected fraction of the data but the larger part of the recorded events. This thesis extends the MUSiC analysis by two components: • Complementary to the standard MUSiC algorithm, a different algorithm is proposed, called “Bump Hunter”. It is a purely data driven analysis, searching for deviations in the smooth distribution of certain observables. As a result, it doesn’t rely on Standard Model Monte Carlo predictions. It is applied to 2010 CMS data. • In order to determine the significance of the deviation between data and Monte Carlo pre- diction, MUSiC calculates a p-value. However, the current implementation shows anoma- lies for certain cases, which have to be corrected for. As a consequence, an alternative p-value will be derived and studied.