I have a dataset that has 17 variables. 9 categorical and 8 continuous. Some have more than 2 levels. I've reduced the dimensionality significantly. I am looking for strategies to test for colinearity within the dataset before I construct the logistic model and test for collinearity there.

I can just split the model into subsets of the categorical and continuous data then test fo collinearity there. Then do so again for the logistic model with an Anova test. But I am not sure what the best options might be.

  • 1
    $\begingroup$ 17 variables is a lot. What is your research question you are trying to address? You should really be able reduce the number of variables down to the ones that address your research question, and then worry about collinearity. EDIT: You should also explain what field you are using the logistic model for. For example, if you are researching econometrics, many variables are sometimes used. Epidemiology, not as much. $\endgroup$ – coconn41 Nov 16 '20 at 20:39
  • $\begingroup$ What is your interest in finding collinearity? $\endgroup$ – Dave Nov 16 '20 at 20:42
  • $\begingroup$ This is for econometrics evaluating loan default risk. The idea is to make a predictive model that can better determine loan default. The logistic model being 16 variables to the target variable of Defaull. Which is binary categorical. This is over 60 years of loan records and several million rows. $\endgroup$ – Randy B. Nov 16 '20 at 23:52
  • $\begingroup$ So what is your interest in finding collinearity? $\endgroup$ – Dave Nov 17 '20 at 2:24
  • $\begingroup$ To see if the predictors have collinearity as I build out the logistic model. $\endgroup$ – Randy B. Nov 17 '20 at 11:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.