[R] lmer output

[Ben Bolker]

 <Denis.Aydin <at> unibas.ch> writes:

> I have a question regarding an output of a binomial lmer-model.
> The model is as follows:
> lmer(y~diet * day * female + (day|female),family=binomial)

  A reproducible example would always be nice.
 
> The corresponding output is:
> Generalized linear mixed model fit by the Laplace approximation 
> Formula: y ~ diet * day * female + (day | female) 
>   AIC  BIC logLik deviance
>  1084 1136 -531.1     1062

[ snip ]

> Fixed effects:
>                     Estimate Std. Error z value Pr(>|z|) 
> (Intercept)         0.996444   0.713703   1.396   0.1627 
> dietNAA             1.194581   0.862294   1.385   0.1659 
> day                 0.142026   0.074270   1.912   0.0558 .
> female              0.015629   0.019156   0.816   0.4146 
> dietNAA:day        -0.124755   0.088684  -1.407   0.1595 
> dietNAA:female     -0.024733   0.026947  -0.918   0.3587 
> day:female         -0.001535   0.001966  -0.781   0.4348 
> dietNAA:day:female  0.001543   0.002720   0.568   0.5704 
> 
> Now from my understanding, the estimates represent differences in slopes 
> and intercepts between different levels of "diet" and so on.
> 
> My questions:
> 
> 1. Is there a way to display the coefficients for all levels of variables 
> (e.g., "dietAA" and "dietNAA")? Because it is quite hard to calculate the 
> slopes and intercepts for all levels of each variable.

See if 
lmer(y~(diet-1) * (day-1) * (female-1) + (day|female),family=binomial)

helps, or see if you can use predict() with an appropriately
constructed prediction data frame -- although not sure if
predict works with GLMMs in current version of lme4.

> 
> 2. Is there a way to get the degrees of freedom?

  Giant can of worms, I'm afraid. See <http://glmm.wikidot.com/faq> for
relevant links and alternatives.