Logistic Regression : Multi Collinearity Check

multi_collinearity
logistic_regression

#1

Dear All,

Would like to know different ways to perform collinearity check while executing logistic regression.

I am aware of some of the ways but certainly not sure of. :smile:

  1. Take each independent variable at a time as a function of other independent variables and calculate VIF. This way we will have an idea of what all independent variables are correlated.

  2. Form a correlation matrix among independent variables and try to figure out the correlations.

  3. Take your dependent and all independent variables and run them through a linear regression with VIF option.

Would really appreciate if anyone can help.

P.S - If someone suggests use of PCA, then do explain the procedure for that as well.

Regards
Saurabh Kapoor


#3

@shuvayan - Thanks for the quick response.

However, the link that you have shared is about validating the model. Apparently, that is not the ask.


#4

@k_saurabh86,

Sorry for the wrong link. :pensive:
I have used VIF option in logistic regression to check for multi - collinearity.If the values are < 10 then there is no problem .
Even the third option that you have mentioned works.


#5

@shuvayan - As far as I am aware, VIF option can’t be used with Proc Logistic. How did you use it? Kindly share the syntax.


#6

@k_saurabh86,

You are right.VIF option is not available for Proc Logistic.You need to run proc reg on the data and just interpret the output of the VIF column.
Note:Other outputs will be non-nonsensical.