[R-sig-eco] NMDS in vegan

[Jari Oksanen]

On Thu, 2008-06-12 at 10:36 -0300, Miltinho Astronauta wrote:
> Hi Manuel,
> 
> As Prof. Jari Oksanen anwered about the metaMDS, permit me comment about the
> usage (or not usage) of NMDS method to reduce the dimensionality of your
> landscape metrics. I think that your are planning keep the firsts dimensions
> for other analyses. Case yes, and in the case you use the axis on
> regression, if need to be care because the axis 1, 2 etc are not ortogonal,
> and the results could be a little bit difficult of explain. There are three
> papers very interesting on this regard of choice (or reduction of the
> number) of landscape metrics:
> 
Howdy Folks,

In fact, the metaMDS ordination scores are orthogonal (because they are
orthogonalized within metaMDS). You can see this with:

cor(scores(mymetaMDSresult)) # or with
crossprod(scores(mymetaMDSresult))

This does not mean that I would recommend using these dimensions in any
regression type analysis: you need all NMDS axes to span the space for
which the solution is optimized, and you can't tear off an axis from the
ordination. That is the reason why vegan has other tools than regression
to inspect the results (ordisurf, envfit).

cheers, jari oksanen
-- 
Jari Oksanen -- Dept Biology, Univ Oulu, FI-90014 Oulu, Finland
email jari.oksanen at oulu.fi, homepage http://cc.oulu.fi/~jarioksa/