3

Before running or building a model, ho can we check on the multicollinearity between different covariates in GLM model in R?

I know that SAS Proc MIXED procedure gives a column for VIF which is very easy but how to check in R?

Stephan Kolassa
  • 95,027
  • 13
  • 197
  • 357
Anjali
  • 891
  • 3
  • 10
  • 10
  • 3
    There are several possibilities to calculate the variance inflation factor. You could try the `vif()` function from the `car` package. – smillig Jul 25 '14 at 11:54
  • 2
    Multicollinearity is a property of the *regressors*, not the *model*, so you don't need to look for "multicollinearity in GLM" as opposed, say, to "multicollinearity in OLS". In addition, there are other measures of multicollinearity than VIF, like the condition indices and variance decomposition proportions of Belsley, Kuh & Welsch, so it would be good if you could edit your question - are you *specifically* interested in the VIF, or *generally* in detecting multicollinearity in R? (I also voted to close and move to stackoverflow.com, since this seems to be specifically about R.) – Stephan Kolassa Jul 25 '14 at 12:28
  • 2
    @Stephan Alternatively, we could take the appearance of this question on CV as a request for information about checking for multicollinearity among regressors (regardless of the computing platform), in which case your comment would be the start of good answer :-). – whuber Jul 25 '14 at 13:15
  • A few words on multicollinearity in general are [here](http://stats.stackexchange.com/questions/3549/f-and-t-statistics-in-a-regression/3557#3557). The [multicollinearity tag](http://stats.stackexchange.com/questions/tagged/multicollinearity) may also be helpful. – Stephan Kolassa Jul 28 '14 at 13:44
  • @Stephen That is correct, but there is a generalization of VIF which does not directly measure collinearity of regressors, but instead analyzes the variance-covariance matrix of the parameter estimates. In particular, the `car:::vif.lm` procedure in `R` implements the GVIF (but calls it "VIF"). Perhaps this will be further elaborated in our new thread at http://stats.stackexchange.com/questions/136004. – whuber Feb 02 '15 at 21:00
  • 1
    @whuber What about a GLMER? Would there also be another way to calculate VIF for it? – Charlie Glez May 26 '16 at 13:44
  • @Charlie I don't know the answer to that. – whuber May 26 '16 at 13:47

0 Answers0