[R] Nested model and variance partitioning

poulet at cict.fr poulet at cict.fr
Thu May 4 15:32:53 CEST 2006

Dear Pierre, thanks for your answer,

I am not quite sure what you mean by "outcome" but I try to be more precise.

I have 4 spatial scales with: 
- One sample per habitat (i.e. substratum)
- Many habitats per site
- One or many sites per river
- Many rivers per regions
- Two regions

Each scales is described by one or many variables:

- Habitat: type of substratum
- Site: altitude, distance from the source, slope, drainage area
- River: drainage area, altitude of the source, slope, type of geology
- Regions: 2 modalities

First I could try a nested ANOVA such as:

SR = habitat + Site / River / Region

But I am rather interested by which variables in each scales explains the
species richness variance. So something like this:

SR = substratum + (alt+dfs+sl+da) / (da+as+s+geol) / Region

But I don't know if i) if it is the good model and ii) how to proceed with the
lmne package.

If someone could give an example close to my case, it will really help me...

Best wishes,

Nicolas Poulet

Selon bady at univ-lyon1.fr:

> hi, hi all,
> > Dear R users,
> > I face to a nested pattern and despite the numerous examples in the help I
> am
> > still confused.
> > I sampled bugs in different habitats within sites which were within rivers
> > themselves within different regions.
> > The habitat correspond to different substrata (not systematically present
> in
> > all sites). For rivers and sites, I have environemental variables (e.g.
> altitude
> > and slope of the site, drainage area and geology of the river) and I have
> > only 2 regions. Note that sometimes I have only one site per river.
> > I would like to know the part of each spatial scale described by
> > environmental
> > data in the species richness variance. For instance is the drainage area
> at
> > the river scale that explains a large amount of species richness variance
> or
> the
> > altitude of the site, or the substratum, etc.
> The design of our data is complex. What is your outcome ?
> > I looked into the nlme package but I did not found how to proceed

> The reference document about the package nlme is certainly “Pinheiro J. C. &
> Bates D. M.  (2000) Mixed-effects models in S and S-plus. Springer Verlag,
> New
> York. 528 pp.”
> but there is many information about this subject on the WEB (see
> http://www.google.fr/search?hl=fr&q=nlme&btnG=Recherche+Google&meta=)
> Regards,
> Pierre
> --------------------------------------------------------------------------
> Ce message a été envoyé depuis le webmail IMP (Internet Messaging Program)

More information about the R-help mailing list