Freiburg 2008 – scientific programme
Parts | Days | Selection | Search | Downloads | Help
T: Fachverband Teilchenphysik
T 68: Experimentelle Methoden
T 68.5: Talk
Friday, March 7, 2008, 15:05–15:20, KGI-HS 1098
BAT - a Bayesian Analysis Toolkit — Allen Caldwell1, Massimo Corradi2, Giulio D’Agostini3, Andrea Knue4, Daniel Kollar1, •Kevin Kröninger4, and Arnulf Quadt4 — 1Max-Planck Institut für Physik, München — 2INFN, Bologna — 3University of Rome La Sapienza — 4II. Physikalisches Institut, Universität Göttingen
The goal of data analysis is to compare models to data and to learn about the validity of these models as representations of the data. In Bayesian inference quantities can be calculated to estimate this validity and to obtain the most suitable set of parameters to describe the data.
The Bayesian Analysis Toolkit, BAT, is a tool developed to evaluate the posterior probability distribution for models and their parameters. The C++ based framework allows a flexible definition of mathematical models and applications. It provides sets of algorithms for numerical integration, optimization and error propagation. Emphasis is placed on the usage of Markov chain Monte Carlo and the possibility to interface BAT to other software such as the CUBA library or Minuit. In addition, methods to judge the “goodness–of–fit” of a model are implemented as well as an interface to ROOT for further analysis and graphical display.
Driven by data analyses in particle physics BAT is being used in a variety of settings. The applications so far range from the extraction of structure functions in ZEUS and the calculation of the sensitivity of GERDA to double beta-decay, to the measurement of the top quark mass with the matrix-element method in ATLAS.