Dear Robert,
> How crucial is it that a given user regressor integrates (over time) to
> zero (ie, zero centered) when putting into a model?
Not crucial but generally it is a good idea to mean correct (i.e.
center) your regressors. This is because they become confounded with
the constant term if not. Although not in itself a problem (i.e. the
parameter estimates and T values will be unaffected) it means that
interactions with this covariate also contain interactions with the
constant (which is the main effect of the other factor).
I hope this helps - Karl
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|