Structured collinearity

[Jihane bou zakhem]

Hello,

Is there any way to remove structured collinearity between two independent variables (x1 and x1*x2), noting that x2 is a dichotomy variable. Please note that centering , transforming to log didn't work..

Thank you,

Jihane

[Mike Lacy]

The problem you think is a problem is almost certainly not a problem. Collinearity is inevitable when a product interaction term is used, and centering does not remove any such "problems." It makes the regression estimates easier to interpret, and it helps people with less knowledge avoid misleading thinking about the "significance" of "main effects." Whether or not you center, the derived slopes and standard errors at various points in your covariate space will be the same. If you post the commands you used and the results you have, and describe why you think there is a problem, one or another member here can explain the issue in the context of your situation. And, by the way, transforming to a log might be useful for various reasons, but it's not a technique to solve collinearity problems.

[Jihane bou zakhem]

Thank you