Automatic Datacard Generation and Significance Estimation with Punzi Criterion for Higgs Analyses
Session Number
Project ID: PHYS 01
Advisor(s)
Dr. Peter Dong; Illinois Mathematics and Science Academy
Discipline
Physical Science
Start Date
19-4-2023 11:55 AM
End Date
19-4-2023 12:20 PM
Abstract
In order to perform a statistical analysis of Higgs searches, datacards are used to encode the observed and expected events after event selection to discriminate between signal and background. We developed a framework to incorporate into our analysis group’s framework to automatically generate datacards. In addition, we utilized the Higgs Combine Tool to test combining multiple generated datacards. In the past, our group’s selection cuts were based on the standard significance = S/sqrt(B) formulation. However, this method is dependent on a-priori expectations of limit settings to assert that a discovery has been made. We instead used the Punzi criterion, developed by George Punzi, as a better estimate of the significance. Simple, approximate formulas were derived and used for Poisson counts with background, with the suggested values for parameters (a=5).
Automatic Datacard Generation and Significance Estimation with Punzi Criterion for Higgs Analyses
In order to perform a statistical analysis of Higgs searches, datacards are used to encode the observed and expected events after event selection to discriminate between signal and background. We developed a framework to incorporate into our analysis group’s framework to automatically generate datacards. In addition, we utilized the Higgs Combine Tool to test combining multiple generated datacards. In the past, our group’s selection cuts were based on the standard significance = S/sqrt(B) formulation. However, this method is dependent on a-priori expectations of limit settings to assert that a discovery has been made. We instead used the Punzi criterion, developed by George Punzi, as a better estimate of the significance. Simple, approximate formulas were derived and used for Poisson counts with background, with the suggested values for parameters (a=5).