This repository contains a tutorial for the zfit and Minuit libraries for parameter estimation in python that was created for the LHCb starterkit 2022. Just click the binder link to try it out!
The lesson is divided into three different parts:
- Part I - zfit Basics sets the scene for the lesson and introduces the basics needed to run a fit in
zfit. The fit result is investigated and different cost functions are introduced. - Part II - zfit Fit Failure tries to give some examples as to why a fit might be unstable or fail and how to avoid them.
- Part III - zfit Advanced shows examples on more advanced use cases for
zfit, like performing pseudoexperiments, statistical background subtraction, simultanous fits, and fits to a disjointed observable space.
The notebook was created having python 3.9 in mind with zfit version 0.10.1 and Minuit 2.17.0.
Additionally required are the dependencies of the packages, such as tensorflow, and numpy.
The way i recommend to set up a virtual environment for python is with miniconda or micromamba. The full requirements can be found in requirements_general.txt. A working requirement for binder is given in requirements.txt.
Many thanks to Jonas Eschle for providing helpful comments when creating this tutorial and for his continued development of zfit!
I also want to thank Lorenzo Paolucci, Gediminas Sarpis, and Dan Thompson for helping me improve the presentation of the content.
Parts of this tutorial are inspired by Hans Dembinskis pyHEP tutorial for Minuit and Statistical Data Analysis by Glen Cowan.