Grama: A Grammar of Model Analysis
Zachary del Rosario · The Journal of Open Source Software · 2020
Grama is a Python package implementing a functional grammar of model analysis emphasizing the quantification of uncertainties.In Grama a model contains both a function mapping inputs to outputs as well as a distribution characterizing uncertainties on those inputs.This conceptual object unifies the engineer/scientist's definition of a model with that of a statistician.Grama provides an implementation of this model concept, as well as verbs to carry out model-building and model-analysis. Statement of NeedUncertainty Quantification (UQ) is the science of analyzing uncertainty in scientific problems and using those results to inform decisions.UQ has important applications to building safety-critical engineering systems, and to making high-consequence choices based on scientific models.However, UQ is generally not taught at the undergraduate level: Many engineers leave their undergraduate training with a purely deterministic view of their discipline, which can lead to probabilistic design errors that negatively impact safety (del Rosario, Fenrich, & Iaccarino, 2020).To that end, I have developed a grammar of model analysis-Grama-to facilitate rapid model analysis, communication of results, and the teaching of concepts, all with quantified uncertainties.Intended users of Grama are scientists and engineers at the undergraduate level and upward, seeking to analyze computationally-lightweight models.