Correlation coeff of logit

[Julie Ann Basconcillo]

Hi!

I have (an) interaction term/s in a logistic regression that I am running. But I always get this above 0.75 coefficient when I check for correlation between the new variable (with interaction) and the original variable using vce, corr. I read that vce, corr is also a test for multicollinearity. This is quite high which is understandable because the interaction term is just a combination of the two. Are the results valid in this case? By the way, I am referring to 2 dummy variables: household head and gender.

Thanks!

[Richard Williams]

This is very common. For a discussion of when and when not to be concerned about collinearity, see

http://statisticalhorizons.com/multicollinearity

[Julie Ann Basconcillo]

Thanks, Richard!