Reference

GUNDAM Documentation

The GUNDAM documentation provides detailed information about the use and applications of GUNDAM, the Generalized and Unified Neutrino Data Analysis Methods: a suite of applications for statistical analyses in neutrino physics, from fitting model parameters to experimental or simulated data to calculating cross sections from the fit results.

A GUNDAM job is set up entirely through YAML or JSON configuration files, so a new study rarely requires touching the code. Most of this documentation therefore explains what each part of a configuration file does and which options it accepts.

Part 1

User Guide

The first part of the documentation introduces GUNDAM and shows how to run its applications.

  • Introduction

    What GUNDAM is, where it comes from, and why it is driven by configuration files, together with the list of GUNDAM releases, each linked to its release page.

  • Run GUNDAM

    How to launch the GUNDAM applications from the command line. It covers gundamFitter, which fits the model parameters to data, and gundamCalcXsec, which calculates cross sections from a gundamFitter output, with their command-line and trigger options. It then describes the helper applications for comparing and unfolding configuration files, reading, comparing and plotting fit results, packaging inputs, and working interactively in ROOT.

Part 2

Configuration Reference

The second part describes every configuration option, following the structure of a configuration file from the top level down.

  • Configuration File Reference

    How to split a configuration into several included YAML files, the top-level options, and the blocks that make up a fit: the fitter engine, the likelihood interface, the propagator, datasets, samples, the plot generator, the event tree writer and the parameter scanner.

  • Parameters and Dials

    How fit parameters are organized into parameter sets, which can share a prior covariance matrix and be eigen-decomposed; how each parameter is defined; and how dials turn parameter values into event weights. Every dial type is covered (normalization, spline, graph, surface, formula, tabulated and kriged), including the external-library interface used by tabulated and kriged dials.

  • Minimizer Configuration

    How to choose between a maximum-likelihood fit with the ROOT minimizer (Minuit2 by default, with HESSE or MINOS errors) and Bayesian posterior sampling with SimpleMcmc, an adaptive Metropolis–Hastings chain, including how to write custom burn-in sequences.

Documentation

Loading…

Loading documentation…