I am very new in r and at analyzing gee models. I have a very high dimensional data (51 independent variables measure at multiple times with no missing values (secondary dataset)). I am pretty sure there is collinearity among the independent variables. So I thought to get the VIF for each predictor using vif from the car package but it comes out this error:vif(fit_full) Error in match.arg(type) : 'arg' should be one of “pearson”, “working”, “response”
My questions are: 1. How to get individual VIF in gee models for each independent variable(my outcome is binary and my independent variables are binary as well)? or how should I assess collinearity in this high-dimensional longitudinal dataset? 2. Once I get the vif, do I remove these high collinear variables before model selection? or what it is the best way to tackle when fitting this gee model?
I am using the geepack in r
Any help would be greatly appreciated. Thanks! Linda