[R-sig-ME] lme vs. lmer and contrasts

[espesser]

In the last discussion about "lme vs. lmer"  (see below) , D. Bates 
adviced to
use anova between 2 mixed models to test a fixed effect.

1) Does it mean that mcmc pvalues have to be disregarded finally ?

2) When the fixed effect is a factor, anova returns the whole 
significance of the factor,
but no information about  the levels of the factor. How is it possible 
to get it  ?
(I am thinking about  contrast analysis for a factor with more than 2 
levels)

3) As a side question:
The mixed models are said to be robust to unbalanced data.
Let suppose a 4 levels factor, with a  number of measures which is 
different for
each level (or at least for one level). In what extend can I trust 
outputs of a mixed model with
typical contrasts (treatment, sdif, ordered) applied to such a factor ?

Thank you very much for your help

R. Espesser
Laboratoire Parole et Langage,
Université de Provence/CNRS,
5 av. Pasteur, Aix en provence (France)


Douglas Bates a écrit :
>
> [...]
>
> My general advice to those who are required to produce a p-value for a
> particular fixed-effects term in a mixed-effects model is to use a
> likelihood ratio test.  Fit the model including that term using
> maximum likelihood (i.e. REML = FALSE), fit it again without the term
> and compare the results using anova.
>
> The likelihood ratio statistic will be compared to a chi-squared
> distribution to get a p-value and this process is somewhat suspect
> when the degrees of freedom would be small.  However, so many other
> things could be going wrong when you are fitting complex models to few
> observations that this may be the least of your worries.
>
>