domir-package {domir}R Documentation

Tools to Support Relative Importance Analysis

Description

Methods to apply dominance analysis-based relative importance analysis for predictive modeling functions.

Details

This package supports the determination of importance for inputs (i.e., independent variables, predictors, features, parameter estimates; called 'names' in the package) using dominance analysis (Azen & Budescu, 2004; Budescu, 1993).

Dominance analysis resolves the indeterminancy of ascribing the value returned by a predictive modeling function to inputs/names when it is not possible to do so analytically. The most common use case for the application of dominance analysis is in comparing inputs/names in terms of their contribution to a predictive model's fit statistic or metric.

Dominance analysis is a common, and generally well accepted, method for determining the relative importance of inputs/names that is, in part, a conceptual extension of the well-known Shapley value decomposition (e.g., Grömping, 2007; Lipovetsky & Conklin, 2001).

Author(s)

Joseph Luchman jluchman@gmail.com

References


[Package domir version 1.2.0 Index]