[R-sig-ME] Questions about using glmmPQL and glmer

[Ben Bolker]

Nai-Wei Chen wrote:
> Dear all R users,
> 
> I  have problems when I use glmmPQL to analyze binary data with
> random effects. When I use "summary(glmmPQL())$residuals", I just see
> the five summary statistics of standardized residuals. How can I
> retrive the fitted probability and reiduals from the summary?

  m1 <- glmmPQL(...)
  residuals(m1)
  fitted(m1)
> 
> When I retrive the fitted values of fixed effects, what do the values
> corresponding to increasing levels of grouping mean?

  See any discussion of contrasts in R: differences between level x and
the baseline (first) level.  (In order to use glmmPQL you should be
familiar with both glm() and lme() ...)
> 
> When I use the "glmer" procedure, how can I retrive the AIC value,
> the coefficients of random effects to the groups and residuals?

  m2 <- glmer()
  AIC(m2) ## or maybe AIC(logLik(m2))
  ranef(m2)
  residuals(m2)

  I would be careful using glmmPQL on binary data, this is a type of
data where penalized quasi-likelihood is known to be a bit dicey ...
Breslow, N. E. 2004. Whither PQL? Pages 1–22 in D. Y. Lin and P. J.
Heagerty, editors. Proceedings of the second Seattle symposium in
biostatistics: Analysis of correlated data. Springer.
<www.bepress.com/uwbiostat/paper192/>