MadAnalysis 5 - Integration of theoretical uncertainty calculation with multiweight integration

Description

MadAnalysis 5 is a framework for phenomenological investigations at particle colliders. Based on a C++ kernel, this program allows to efficiently perform, in a straightforward and user-friendly fashion, sophisticated physics analyses of event files such as those generated by a large class of Monte Carlo (MC) event generators. MadAnalysis 5 comes with two modes of running. The first one, easier to handle, uses the strengths of a powerful Python interface to implement the analysis utilizing a set of intuitive commands. The second one requires implementing the analysis in the C++ programming language directly within the core of the analysis framework. This opens unlimited possibilities concerning the level of complexity that can be reached by the analysis, which is only limited by the programming skills and the originality of the user.

Monte Carlo is a method for sampling high-dimensional parameter spaces. The method requires knowledge of the weight function determining the probability that a state is observed. Due to the various unknowns in the event generation process, a collision event generated by an MC event generator carries various uncertainties, i.e. scale and PDF uncertainties. Currently, MadAnalysis 5 framework can only use constant uncertainty values for the entire event sample, which is then used in the reinterpretation of the exclusion limits of a given BSM model (for details see this study). This project will focus on generalizing the uncertainty module within MadAnalysis 5, where instead of using overall uncertainties for a given sample, each event will be assigned with an uncertainty value corresponding to its phase-space. This will allow MadAnalysis 5 framework to mitigate the phase space information through uncertainties and calculate more accurate exclusion limits.

Task ideas

This project will involve the redesign of the current data structure for event weights. MadAnalysis 5 is capable of reading all the weights presented in an MC event sample; however, these uncertainties are not mitigated to cut-flow and histogram classes where these modules are only executed with the nominal weight of the event. The new data structure design of weights will be integrated with cut-flow and histogramming modules to include the complete uncertainty information while preserving the backwards compatibility of the current interface.

Expected results & Milestones

Requirements

Mentors

Additional Information

Corresponding Project

Participating Organizations