[ADMB Users] glmmADMB in R:question regarding 'residuals(Model)'

[H. Skaug]

Hi,

I think residuals(best) returns

[A1$NO -(fitted(best)] / SD

where SD is the standard deviation, which depends on the distribution
at hand. The code
inside glmm.admb that determines SD is:

   tmpsd <- switch(family, poisson = sqrt(lambda), nbinom = sqrt(lambda *
        (1 + lambda/out$alpha)), binomial = sqrt(out$fitted *
        (1 - out$fitted)))


Hans


On Fri, Jan 7, 2011 at 1:01 AM, Schweizer, Peter E.
<schweizerpe at ornl.gov> wrote:
> Dear colleagues,
>
>
>
> We are using glmmADMB in R to model land cover and water quality influence
> on species diversity of fishes within a study area with several subregions.
>
> We defined subregion as a random factor and also ask for individual
> intercepts for the different subregions.
>
> A ‘global’ model for overdispersed count data was formulated as
>
> GM<-glmm.admb( N_Species ~ b1 + b2 + b3 + …+ bn + Subregion, random = ~ 1,
> group="Subregion", data=input, family="nbinom")
>
> We subsequently evaluated several candidate models that represent various
> subsets of variables from the global model.
>
> Our input file is A1, with A1$NO representing the observed number of
> species. During the process of examining model performance we used
>
> Observed – Predicted   (A1$NO -(fitted(best)))  for the ‘best’ model based
> on lowest AICc to derive residuals for predicted Nspecies. However, using
> ‘residuals(best)’ produced considerable different (smaller) values which we
> find somehow puzzling. Are we wrong to assume that (Nspecies predicted by
> ‘best’ model,  + residuals(best)) should add up to Nspecies observed
> (A1$NO)?
>
>
>
> Any insight would be appreciated.
>
>
>
> Peter
>
> _______________________________________________
> Users mailing list
> Users at admb-project.org
> http://lists.admb-project.org/mailman/listinfo/users
>
>