At what point does the correlation coefficient between two variables rule out the inclusion of one of those variables as a regressor on the other in multiple regression? Is between .6-.7 too high?

As I understand it, this is a distinct question from (multi)collinearity statistics, which answers this question with regard to the relationship between all predictor variables (not predictors and the dependent).

Many thanks in advance for any help you can offer!

  • $\begingroup$ In what context? There are applications of regression where correlation of predictors is irrelevent, and applications where it is very relevent, any many degrees in the middle. $\endgroup$ – Matthew Drury Jun 25 '17 at 0:21
  • $\begingroup$ Hierarchical linear regression with sample size(s) ranging from 700-1200. Dependent variables are individual difference measures and predictor variables are individual differences and demographics. There will be 3-5 predictor variables in the regression analysis as well. Does this help ? $\endgroup$ – brainman Jun 25 '17 at 15:59
  • $\begingroup$ A bit. More important is what you are building the regression to accomplish. Are you interested in prediction, inference, both? $\endgroup$ – Matthew Drury Jun 25 '17 at 18:07
  • $\begingroup$ Both, but predominantly, and more conservatively, to predict scores in this data set. It is an entirely theory driven analysis and set of experiments with very clear, distinct and falsifiable predictions. $\endgroup$ – brainman Jun 25 '17 at 23:01

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.