Thursday, November 20, 2008

Based on the last question:

Can we use a calculated interaction term for question 1 with gender, by treating gender as continuous?

Furthermore, can we center gender if we have problems with collinearity? Is it okay to center gender?

1 comment:

Mari said...

Please refer to the slide & text on dichotomous variables in moderation to determine what your options are for examining gender.

Variables that are coded 0 & 1 are not typically centered, because the Y intercept would no longer make any sense, and this is the primary reason for centering a variable.

All moderator and predictor variables that are not contrast coded should, however, be centered before creating an interaction term.