[R-sig-ME] Attempts at diagnostics with logistic regression and two random effects

Mitchell Maltenfort mmalten at gmail.com
Tue May 25 21:23:18 CEST 2010

I am currently playing with a logistic regression (family=binomial),
two random effects (state and hospital within state), and a handful of
fixed effects.

I'm implementing the two random effects as 1|hospital + 1|state,
following Dr. Bates' draft chapter 2 for the lme4 book.

So far, I've found that plotting prediction vs. residual gives me two
curved lines,  both concave facing the right (like the reciprocal
curve for x > 0), and that doing qqnorm on the residual gives me an
approximate straight line that then shoots upward.

Not quite sure how to interpret what they are telling me, never mind
addressing the inherent problem.

Can anyone point me in the right direction?


More information about the R-sig-mixed-models mailing list