Dear Jeanette,
thanks very much for your comments. I looked at the website several times, it's a great reference. However I am not sure if it generalizes to several covariates. I have demeaned age already and have set up the model except demeaning acquisition, which I'll do now. So coming back to generalizing: I understand that it won't matter for the interaction whether I demean or not, only if I'd look at the group effects as well. Sorry that's probably quite repetitive, but I can demean all covariates, ie demeaning one will have no influence on the other being demeaned as well? As I am only taking the mean out of either covariate I'll keep relative differences - I am almost certain that's how it is :)
Regarding reverting back to a simpler model in the absence of an interaction, I'd keep the acquisition covariate though?
Thanks for your help!
Best,
Torsten
|