varimp.logistic: A function to report variable importance after glm() or...

In fzettelmeyer/mktg482: Functions needed for MKTG 482

Description

This function reports standardized coefficients and ranks variable by importance: The coefficients of continuous variables are standardized to a two standard deviation change of the variable. The coefficients for factor variables are left unchanged. This follows the procedure suggested by Andrew Gelman in "Scaling regression inputs by dividing by two standard deviations," Statistics in Medicine (2008), Vol. 27, pp. 2965-2873. The function takes as inputs models created by glm or caret using glm

Usage

varimp.logistic(modelFit)

## S3 method for class 'glm'
varimp.logistic(modelFit)

## S3 method for class 'train'
varimp.logistic(modelFit)

Arguments

modelFit

reguired: estimated/trained glm model

Value

A tibble with variable, var_imp, p_value, factor, OR_std, OR_sd_perc

Methods (by class)

• glm: Method for glm()

• train: Method for either glm or glmnet in caret. This works regardless of whether the data is pre-processed with "scale"

Examples

varimp.logistic(logit1)
varimp.logistic(logitFit)

fzettelmeyer/mktg482 documentation built on May 26, 2020, 2:26 p.m.