I do not believe in centering categorical variables because it
doesn’t make sense to consider the slope and intercept when the gender, say, is
halfway between male and female. Center the continuous variable and leave the
dichotomous coded 0,1. I wouldn’t worry about collinearity unless it is really
bad, which would indicate substantial difference in the predictor by gender.
Dr. Paul R. Swank,
Professor and Director of Research
Children's Learning Institute
University of Texas Health Science Center-Houston
From: SPSSX(r) Discussion
[mailto:[hidden email]] On Behalf Of Handel, Richard W.
Sent: Wednesday, May 06, 2009 9:02 AM
To: [hidden email]
Subject: regression with an interaction term (centering predictors)
Hi All,
I have a question about centering predictors in a regression
that includes a dichotomous variable. How should one handle the
dichotomous variable? Say you have a continuous variable, a gender variable (0
or 1), and an interaction between the continuous variable and the gender variable.
If I just center the continuous variable prior to calculating the interaction
term, I noticed
that some of the collinearity indices actually look worse
than not centering at all. Are you supposed to center the dichotomous variable
as well (this seems counterintuitive to me) or are you not supposed to center
any variables if your analysis includes a dichotomous variable?
Thanks!!!
Rick
Richard W. Handel, Ph.D.
Associate Professor
Department of Psychiatry and Behavioral Sciences
Eastern Virginia Medical School
825 Fairfax Avenue, Hofheimer Hall
Norfolk, VA 23507
Phone (757)-446-7992
Locked
