Speaker
Wouter Verkerke
(NIKHEF (NL))
Description
RooFit is a library of C++ classes that facilitate data modeling in the ROOT environment. Mathematical concepts such as variables, (probability density) functions and integrals are represented as C++ objects. The package provides a flexible framework for building complex fit models through classes that mimic math operators. For all constructed models RooFit provides a concise yet powerful interface for fitting, plotting and toy Monte Carlo generation as well as sophisticated tools to manage large scale projects.
RooFit has been used in countless published B-factory and LHC results. We will review recent developments such as the ability to persist models in ROOT files in container classes, which enables the concept of digital publishing of analytical likelihood functions with an arbitrary number of parameters. Persistent models enable completely ways to perform physics analyses: Complex fit results can be trivially shared between groups and experiments for validation and detailed combinations of physics results, such as the combination of Higgs decay channels, can be constructed in a matter of hours. Finally, model persistence simplifies the streaming of tasks to other hosts to parallelize calculation of computing intensive problems that are common in statistical techniques.
RooStats is a project providing advanced statistical tools required for the analysis of LHC data, with emphasis on discoveries, confidence intervals, and combined measurements in the both the Bayesian and Frequentist approaches. The tools are built on top of the RooFit data modeling language and core ROOT mathematics libraries and persistence technology. These tools have been developed in collaboration with the LHC experiments and used by them to produce numerous physics results, the discovery of the the Higgs boson by ATLAS and CMS Higgs, using models with more than 1000 parameters. We will review new developments which have been included in RooStats and the performance optimizations, required to cope with such complex models used by the LHC experiments. We will show as well the parallelization capability of these statistical tools using multiple-processors via PROOF.
Author
Wouter Verkerke
(NIKHEF (NL))
Co-authors
Gena Kukartsev
(Brown University (US))
Giovanni Petrucciani
(Univ. of California San Diego (US))
Dr
Gregory Alfred Schott
(KIT - Karlsruhe Institute of Technology (DE))
Kyle Stuart Cranmer
(New York University (US))
Lorenzo Moneta
(CERN)
Sven Kreiss
(New York University (US))
