1. When conducting a hierarchical multiple regression, to check for multicollinearity do you enter all of your IVs at once or do you enter them in their respective blocks?
  2. If you enter them in their respective blocks and your VIF and Tolerance are okay, but your condition index for the last dimension is over 30, what do you need to look at in terms of the variance proportions to determine if multicollinearity is in fact an issue

The hierarchical formulation of the model is a way of organizing it conceptually but, if you write out the whole model, they're all just treated as predictors. You could have cross-level collinearity that presents a problem. For example, if it were person-within-school, demographic segregation by school district could cause large cross-level correlations.


1) do not just check within level

2) In the situation you described, it sounds like collinearity is an issue. You might want to delete the predictor with the huge condition index.

  • $\begingroup$ Thank you for your help. When I run the predictors simultaneously the condition index is not a problem. However, I'm slightly confused as to what exactly the "dimensions" mean, because it appears that it is the final dimension that is consistently the problem. I had read somewhere that if there is an overlap with only one predictor and the constant then you're okay (if both proportion variances are high), it only becomes an issue if two of the predictors have high proportion variances. Thoughts? $\endgroup$ – Lauren Mar 9 '17 at 14:17

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.