I am running a mediation model with 2 interaction terms (gender x happiness and gender x hours spent working per week) and one dependent term (income). I am wondering if I should mean center these variables before I run the mediation model? Thanks!
4
-
$\begingroup$ Why do you want to center them? $\endgroup$– user2974951Jul 12, 2019 at 5:48
-
$\begingroup$ I'm concerned about multicollinearity. $\endgroup$– SarahJul 12, 2019 at 14:13
-
$\begingroup$ Centering could help with multicollinearity. Of course you can always center the variables, the model will work the same, the only thing that changes is the interpretation of the results. You should still check for multicollinearity, for ex. using VIF metric. $\endgroup$– user2974951Jul 12, 2019 at 18:35
-
$\begingroup$ Centering doesn't help with multicolinearity. The precision of predicted values will be the same regardless of whether you center or not. It only changes the interpretation of the effects. The standard errors will change, but only because you're estimating a different an effect than you would be if you were to use the un-centered variables. $\endgroup$– NoahJul 17, 2019 at 16:16
Add a comment
|