Hi Henrike
> Hi all,
> We analyzing data from a multicenter study. To analyze data at a group level, we included an additonal covariate to model the effect "center". Each subject from center 1 got a "1", from center 2 a "2" and so on. Is that the right way? Or does it modell the center effect in a linear way (that would not be good, of course). Would it then be better to include a new factor ("center") with 4 steps?
> Can someone help?
>
I'll try but no warranty I'm right ..
I think that adding a covariate isn't a good solution - adding a
regressor from 1 to 4 means that you look for an additional variation
which increases from center 1 to 4 and then remove this effect
(covariate). What I suggest is instead to model the centers as a full
variable with 4 levels and use contrasts to pull together the conditions
you want.
Hope this makes sense to you ...
Best
Cyril
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.
|