myFitter documentation

These pages describe the myFitter package - a Python module for global fits and likelihood ratio tests. Its features include:

• computing maximum likelihood estimates of model parameters,
• handling of non-linear constraints on the parameter space,
• computation of profile likelihoods,
• visualisation of one and two-dimensional confidence regions (via matplotlib),
• numerical computation of p-values in likelihood ratio tests by means of importance sampling (myFitter method),
• parallelisation.

The concepts and algorithms implemented in myFitter are published in

M. Wiebusch, “Numerical Computation of p-values with myFitter”, Comput. Phys. Commun. 184 (2013) 2438, doi:10.1016/j.cpc.2013.06.008, arXiv:1207.1446v2

If you use the myFitter package for a scientific publication, please cite this paper.

Contents:

• Getting started
  • Installation
  • Examples
• Statistical inference
  • General case
  • Linear regression models and Wilks’ theorem
  • Plug-in p-values and the myFitter method
  • Gaussian and systematic errors
• The Model class
  • Chi-square contributions
  • Dependent quantities and observables
  • Deriving the model class
  • Gaussian nuisance parameters
  • Manipulating Model objects
• Fitting the model
• Profiling the likelihood
  • One-dimensional profiles
  • Two-dimensional profiles
  • Parallelisation
  • Writing your own profiler
• Toy simulations
  • Generating toy observables
  • Computing p-values
  • The myFitter method
  • Filling histograms
• Class reference
  • Parameters
  • Constraints
  • Chi-square contributions
  • Models
  • Fits
  • Conversions for p, Z, and chi-square values
  • Profilers
  • Gaussian variates
  • Histograms
  • Integration and toy simulations
  • Parallelisation

Indices and tables

• Index
• Search Page