To determine whether multiple independent variables predict a dependent variable, I'm using a hierarchical model of three levels (first level: only control variables, second: control and independent variables, third: control, independent and an interaction variable).

During research online, there appears to be a diverse opinion on whether or not to center the variables. However, a often recurring argument is to perform centering for the interaction term.

I thus centered my interaction variable (using the mean) but this resulted in my model to be only significant on level 1. Nonetheless, before I started centering this interaction term, the model was significant on levels 1 and 3 (i.e. the independent variables [level 2] did not explain the model until a moderator was considered [level 3]).

This is quit annoying as it would seem that the analysis with the uncentered interaction term yields a more interesting relation. Thus, is it possible that centering an interaction term can result in complete different conclusions? Or am I doing probably something wrong?

  • $\begingroup$ Did you center before you created the interaction term or after? $\endgroup$ – gung Apr 16 '16 at 22:20
  • $\begingroup$ @gung Before creating the interaction term $\endgroup$ – Bensa Apr 16 '16 at 22:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.