Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
T: Fachverband Teilchenphysik
T 78: Experimentelle Methoden I
T 78.6: Vortrag
Montag, 28. März 2011, 18:00–18:15, 30.23: 2-1
BAT - Bayesian Analysis Toolkit — Frederik Beaujean1, Allen Caldwell1, Daniel Kollár2, Kevin Kröninger3, •Shabnaz Pashapour3, and Arnulf Quadt3 — 1MPI für Physik, München, Germany — 2CERN, Geneva, Switzerland — 3Georg-August-Universität Göttingen, Göttingen, Germany
One of the most vital steps in any data analysis is the statistical analysis and comparison with the prediction of a theoretical model. The many uncertainties associated with the theoretical model and the observed data require a robust statistical analysis tool.
The Bayesian Analysis Toolkit (BAT) is a powerful statistical analysis software package based on Bayes’ Theorem, developed to evaluate the posterior probability distribution for models and their parameters. It implements Markov Chain Monte Carlo to get the full posterior probability distribution that in turn provides a straightforward parameter estimation, limit setting and uncertainty propagation. Additional algorithms, such as Simulated Annealing, allow to evaluate the global mode of the posterior.
BAT is developed in C++ and allows for a flexible definition of models. A set of predefined models covering standard statistical cases are also included in BAT. It has been interfaced to other commonly used software packages such as ROOT, Minuit, RooStats and CUBA.
An overview of the software and its algorithms will be provided along with several physics examples to cover a range of applications of this statistical tool. Future plans, new features and recent developments will be briefly discussed.