PhD defence by Ask Emil Løvschall-Jensen

Search fore new physics in multilepton final-states using multivariate techniques

A novel use of multivariate methods has been developed for searches for new physics in high-energy particle physics. Through the use of Principal Component Analysis a range of input kinematic and event-observables are transformed into linear uncorrelated principal components. The resulting principal components are ranked by variance associated to the degree of information they contain. The potential usage of this transformation on both data and model expectations is presented and it is shown how the highest ranked principal component can be used in a composite likelihood fit of data to model expectation, combining much more of the available information than any one original observable could provide.

The independence of the principal components is tested using the Mutual Information measure. Principal components found to be independent are combined to add an even larger amount of information to the fit and improve separation power to new physics. This is applied in both tests of the Standard Model and a search for new physics. As a benchmark model upper limits are set on the Type III seesaw mechanism.

The method is shown to hold significant potential for use in both searches for new physics and in testing the Standard Model.