domir-package {domir}R Documentation

Tools to Support Relative Importance Analysis


Methods to apply decomposition-based relative importance analysis for R functions.


Determining the relative importance of inputs to (i.e., independent variables, predictors, features) to a predictive model is topic of interest to scientists and analysts. Decomposing a returned value, such as a model fit metric or statistic, into parts attributable to each input is a commonly applied method for determining relative importance in predictive models.

This package supports applying decomposition methods using lapply- or Map-like functions that compute dominance analysis (Azen & Budescu, 2004; Budescu, 1993)/Shapley value decomposition (Grömping, 2007; Lipovetsky & Conklin, 2001) based on the values returned from other, predictive modeling, functions.

The user interface is structured such that domir automates the decomposition of the returned value and comparisons between model inputs and the user provides the analysis pipeline including model inputs, the predictive modeling function into which they are entered, and returned value from the model to decompose.

This package's user interface accepts inputs as names on the right hand side of a formula which can be passed on to the predictive model directly or further processed in the analysis pipeline. The interface is also planned to be extended to Formula from the package {Formula} as well as list types as inputs.


Joseph Luchman


[Package domir version 1.0.0 Index]