I have a question concerning how to handle multicollinearity in PLS:
One of the ways to help handle multicollinearity in multiple regression is to regress one independent variable onto another independent variable to try to parcel out unique effects.
Can I do something similar in PLS by drawing a line from one independent variable to another? I notice that when I do this the r-squared values in the dependent variables do not change, but several more beta weights leading to the dependent variables now become significant.
If I were running multiple regression, I would see this as a sign that I have at least partially accounted for multicollinearity. Is it the same principle in PLS?
PS I already did choose to standardize the variables in the PLS Algorithm dialogue box.
Multicollinearity question
Multicollinearity question
Michael Palanski
Rochester Institute of Technology
Rochester Institute of Technology