How to test multicollinearity in multinomil logistic regression? I have 25 independent variables and 1 dependent variable. Out of 25 independents variables, 17 variables are continuous variables and 8 are categorical (having two values either Yes/No OR sufficient/Insufficient). I want to check multicollinearity among these independent variables. I am using R. Thanks in advance!
0
votes
1 Answers
0
votes
I believe assessing multicollinearity with a multinomil logres is no different than a standard glm. Many methods to do this are available in R.
One example is:
fit = `your model`
library(car)
vif(fit)
The vif functions stands for variance inflation factor. When you have collinearity variance is increased with little increase in model fit. This function measures this. You want to look at the third column of the output and identify any variables that differ substantially from 1.
The above approach looks at both categorical and continuous predictors.
