[R-sig-ME] lme interpretation

[Ben Bolker]

Joanie VanDeWalle <r4inbow_ at ...> writes:

> 
> 
> Hi,
> 
> I wrote a lme that goes like this:
> 
> f9 <- lme(masse ~ jour + annee + site + sexe + site*annee, 
> random=~1+jour|phoque, data=data3)

  you don't need to add 'annee+site' to this model,
they are implicit in 'site*annee' (which expands to
'site+annee+site:annee'
  
> masse : seal mass, continuous
> jour : seal age at capture (days); numeric and continous (range from 1 to 34)
> annee : year where data was collected, factor of 4 levels (2008-2011)
> site : capture site, factor of 2 levels 
> sexe: seal sex, factor of two levels (male, female)
> phoque: seal individual (random effect)
> 
> There is two things I want to know:
> 
> First, I want to know wether any of the parameters
> have significant effects and the importance of these
> effects. I already did this and selected the model with 
> the lowest AIC. This model is the one I wrote above.
> The second and the most important point is that I 
> want to obtain some "basic" informations about my dataset.
> I want to use my model to determine certain generalities like:
> 
> What is the growth rate (slope of the regression)
>  of the males in general (in terms of mean and SE)?
> What is the birth mass (mass at day 1) of the females 
> in 2010 (also in terms of mean and SE)? 
> 
> Is there a special command in R that could help me do that?
> 

   In this case, since you have not included any interactions of
growth rate (effect of 'jour') with other predictions, coef(f9)["jour"]
should give you the growth rate for *all* individuals, and
summary(f9)["jour",] should give you the mean/SE etc.

  For other examples (those not represented as a single
parameter in the model), see ?predict.lme , and code on
http://glmm.wikidot.com/faq for getting SEs (although *note* 
that the recipe given there ignores uncertainty in 
the estimates of the random effects
variances).

  In some cases, the answer may depend on how you want to average
across the other predictors in the model.