[R-sig-ME] Adjusting for baseline differences in growth models
ken at kjbeath.com.au
Fri Sep 5 00:56:31 CEST 2008
On 03/09/2008, at 1:34 PM, D Chaws wrote:
> Hi all, I posted a long question for this the other day, but no one
> has responded. Here is a short version those of you who are very busy
> might have time to read. Any advice would be greatly appreciated.
> I have a growth model where 2 groups differ on the outcome at
> baseline. How do I get the growth model adjusting for this difference
> (or making the groups equivalent at baseline, such as in ANCOVA)?
> Predict growth from a 2nd(person)-level covariate of baseline scores
> (i.e., a growth:baseline covariate interaction in a mixed model)?
> Does it make sense to add the baseline covariate as a predictor of the
> intercepts as well as the slopes or is this completely redundant with
> a model where time is coded so that 0 = baseline?
Fitting a mixed-effects model for the post baseline outcomes only with
the baseline as a covariate affecting only the intercept seems
sensible. This is the same as the ANCOVA but with repeated measures.
> Thanks again.
> - DC
> R-sig-mixed-models at r-project.org mailing list
More information about the R-sig-mixed-models