Constrained fits with non-Gaussian distributions

17th International Workshop on Advanced Computing and Analysis Techniques in Physics Research

物理学;计算机科学

Frühwirth, R.^1 ; Cencic, O.^2

Institute of High Energy Physics, Austrian Academy of Sciences, Vienna, Austria^1

Institute of Water Quality, Resources and Waste Management, TU Wien, Vienna, Austria^2

关键词: Bayesian reasoning;    Data reconciliation;    Distributed data;    Event reconstruction;    Material flow analysis;    Non-gaussian distribution;    Non-linear constraints;    Outlying observation;   

Others  :  https://iopscience.iop.org/article/10.1088/1742-6596/762/1/012037/pdf
DOI  :  10.1088/1742-6596/762/1/012037

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

Non-normally distributed data are ubiquitous in many areas of science, including high-energy physics. We present a general formalism for constrained fits, also called data reconciliation, with data that are not normally distributed. It is based on Bayesian reasoning and implemented via MCMC sampling. We show how systems of both linear and non-linear constraints can be efficiently treated. We also show how the fit can be made robust against outlying observations. The method is demonstrated on a couple of examples ranging from material flow analysis to the combination of non-normal measurements. Finally, we discuss possible applications in the field of event reconstruction, such as vertex fitting and kinematic fitting with non-normal track errors.