All Times ET
Virtual
Applying Relative Importance Analysis to Models in R: Domir (305315)
*Joseph Luchman, Fors Marsh Group LLC
Keywords: R, Relative Importance, Dominance Analysis, Programming
Determining the relative importance of independent variables in statistical models is a topic of interest to practicing data analysts to evaluate a model and contextualize its predictions given the data. This poster presents on a relatively new R package that permits applying a popular method, dominance analysis (DA; Budescu, 1993), to effectively any statistical or machine learning model. The {domir} package provides a function that can apply the DA evaluation approach to models that use R’s formula interface (or can be adapted to do so by the user) which opens the methodology up to nearly any modeling function in the R computing environment. The focus of this poster will be on describing the building blocks of the package as a lapply- or Map-like function focusing on computing all model subsets. The poster will provide data analytic examples using several R modeling functions and comparisons of capabilities to existing R packages.