Regression Discussion > Centering variables
It is true that centering variables prior to including them in an interaction term will help reduce multicollinearity. This should be done BEFORE creating your interaction term (the centered variables should be used to create the interaction term).
To mean center a variable (the type of centering you'd most likely want to use here), simply subtract the mean of that variable from each person in your dataset and create a variable from the resulting values. The syntax would be basically as follows:
COMPUTE VAR.C = VARIABLENAME - ####.
EXECUTE.
In the syntax above, VAR.C is an arbitrary name that I gave the centered variable, because I like to use the original variable name + ".c" in my centered variable names. The "VARIABLENAME" should be replaced with the name of the variable you are trying to center and #### should be replaced with the mean of the variable you are trying to measure. After you do that process with both variables involved in the interaction, you may create your interaction term.
I hope this helps!
Jeremy Taylor
I am running a multiple regression (MR) for my master's thesis. My variables needed to be transformed due to severe violations of normality. My MR also includes interaction terms. My professor tells me that centering is needed before creating interaction terms to avoid multicollinearity. Is this correct? Does one center the transformed variables, or is centering done on the raw variables in lieu of a conventional transformation? And what are the steps to centering the variables (assuming I have to) in SPSS?