[R-sig-ME] fixed vs random

[Stuart Luppescu]

On Sun, 2010-03-28 at 13:21 -0700, Iasonas Lamprianou wrote:
> m1<- lmer(score ~ 1+(1+gender|candidate), mg2006_sub)

And the other unusual thing about this model, as was pointed out by 
Doug Bates recently in another thread, is that without a fixed effect of
gender, you're stipulating that the overall effect of gender is 0,
leaving only the random effect. In this case, it might not be terrible
(assuming gender is coded {0, 1}) but it might not be what you want,
either.
-- 
Stuart Luppescu <slu at ccsr.uchicago.edu>
University of Chicago