R help - Jul 2012 - Mixed-model with paired design & covariates

Dear all,

I make habitat suitability models for animal species. The purpose of my
research is to investigate the accuracy of different models.

I clearly have a nested design:

- accuracy_measure -> response variable
- 2 model types (model_type) -> fixed effect
- 230 species (species) -> random effect
- 10 replicates/species (replicate) -> random effect
- 10 subreplicate/replicate -> observation

So I have: 10*10*230 observations/model, identified as
speciesID_replicate_subreplicate (species_ID ranging 1:230, replicate 1:10
and subreplicate 1:10)

One could think about such mixed-effect model:

my.model<-lme(fixed = accuracy_measure~model_type, random
~1|species/replicate)

I do not expand here into model simplification nor if it is best to use lme
or lmer, YET... but here are more conceptual questions

1) my replicates & subreplicates are paired in the sense that they come
from the same split of the data. As an example for species X, the 20
observations of replicate1 of model A and B (10 for A and 10 for B) are
linked by a same data split which is likely to influence my accuracy
measure. In the same way the 2 observations replicate1_subreplicate2 of
model A and B (1 for A and 1 for B) are also linked. Is there a way to
introduce such pairing in a mixed effect model?

2) several continuous covariates, attributes of species (number of points
for modelization, size in km2 of the range) may influence the measures of
accuracy and I may be interested in investigating those effects. How could
I include a covariate in such model? How should I strucutre it given that
there are species covariates (high order of nesting)?

I hope that my questions are relevant!
Thank you very much in advance for your help!


Nathan

	[[alternative HTML version deleted]]

Wrong list! Post on r-sig-mixed-models instead of here.

However, this is really not an R question. It appears that you are
looking for remote statistical consulting, which I consider rather
hazardous. Sort of like having a virtual thesis advisor. I believe you
would be better off talking with your local statistician/statistical
expert.

-- Bert

On Sat, Jul 28, 2012 at 4:58 AM, Nathan Ranc <nathan.ranc at gmail.com>
wrote:
> Dear all,
>
> I make habitat suitability models for animal species. The purpose of my
> research is to investigate the accuracy of different models.
>
> I clearly have a nested design:
>
> - accuracy_measure -> response variable
> - 2 model types (model_type) -> fixed effect
> - 230 species (species) -> random effect
> - 10 replicates/species (replicate) -> random effect
> - 10 subreplicate/replicate -> observation
>
> So I have: 10*10*230 observations/model, identified as
> speciesID_replicate_subreplicate (species_ID ranging 1:230, replicate 1:10
> and subreplicate 1:10)
>
> One could think about such mixed-effect model:
>
> my.model<-lme(fixed = accuracy_measure~model_type, random >
~1|species/replicate)
>
> I do not expand here into model simplification nor if it is best to use lme
> or lmer, YET... but here are more conceptual questions
>
> 1) my replicates & subreplicates are paired in the sense that they come
> from the same split of the data. As an example for species X, the 20
> observations of replicate1 of model A and B (10 for A and 10 for B) are
> linked by a same data split which is likely to influence my accuracy
> measure. In the same way the 2 observations replicate1_subreplicate2 of
> model A and B (1 for A and 1 for B) are also linked. Is there a way to
> introduce such pairing in a mixed effect model?
>
> 2) several continuous covariates, attributes of species (number of points
> for modelization, size in km2 of the range) may influence the measures of
> accuracy and I may be interested in investigating those effects. How could
> I include a covariate in such model? How should I strucutre it given that
> there are species covariates (high order of nesting)?
>
> I hope that my questions are relevant!
> Thank you very much in advance for your help!
>
>
> Nathan
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 

Bert Gunter
Genentech Nonclinical Biostatistics

Internal Contact Info:
Phone: 467-7374
Website:
http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm