2
$\begingroup$

I put this question because while reading the benefits of standardizing explanatory variables or not, I read good but contrasting opinions about standardizing when there are interaction in the model.

Some talk about how problems of collinearity are removed when standardizing (e.g. Collinearity diagnostics problematic only when the interaction term is included), which is basically the case of my GLMM. However, others claim that standard errors and p-values of interactions of standardized models are not reliable... (e.g.Variables are often adjusted (e.g. standardised) before making a model - when is this a good idea, and when is it a bad one? or http://quantpsy.org/interact/interactions.htm)

So, any ideas on what is the right thing to do?

$\endgroup$
  • 2
    $\begingroup$ Welcome to Cross Validated! Please have a look at the possible duplicate & if you're still in doubt edit your question to explain how. $\endgroup$ – Scortchi - Reinstate Monica Oct 10 '13 at 20:19
  • 2
    $\begingroup$ "standard errors and p-values of interactions of standardized models are not reliable"? That is not true as far as I know. SE and p-values are equivalent between standardized and un-standardized models. $\endgroup$ – Affine Oct 10 '13 at 20:25

Browse other questions tagged or ask your own question.