three way interactions and centering in multiple regression

Hello, I am examining three variables to see whether they interact with each other to predict my DV using multiple regression. I typically center my variables to make interpretation easier. I was wondering, if my three IVs (let's call them X, Y, and Z) are all centered, then when creating the interaction terms (e.g., XY, XZ, YZ, and XYZ), do I use the uncentered or centered variables? Also, how does one go about probing the three way interaction if significant? Any info would be great! Thanks!


TS Contributor
i see no reason to use interaction. Also why did you center the variables in the first place? Did you have any evidence of doing so because it is more difficult to interpret it that way.
centering and three way interactions

My main question of interest is whether these three IVs interact with each other to predict the DV. I was taught to center variables as this makes interpretation easier (Aiken & West, 1991). I just wasn't sure if, when creating the interaction terms do you use the uncentered or the centered variables?


TS Contributor
The centered ones of course but you may still have problem. I did it with some data and there was still some problem. Your IV they are continuous right?