[R] Confusion metaMDS and CA

[Gavin Simpson]

On Fri, 2009-09-04 at 03:36 -0700, swertie wrote:
> Thank you very much. I am just concerned because I wonder if I used the best
> method. I have presence/absence data. With isoMDS I can specify
> "Bray-Curtis" distance method, which is adequate, but I was not sure of the
> method used by metaMDS. I think that it is Euclidian distance. Can I use it
> for presence/absence data? 

metaMDS calls isoMDS internally so whatever you do with isoMDS can be
done with metaMDS. metaMDS adds some nice sugar so you don't have to
deal with running several random starts and evaluate the solutions, and
it also adds some features that have been shown to be useful for
ecological applications.

Check out Minchin (1987) paper cited in the references section
of ?metaMDS for one evaluation of nMDS versus other ordination
techniques.

The default dissimilarity coefficient used in metaMDS is the
bray-curtis, not the Euclidean as you thought. But you can use any of
the dissimilarities within the function vegdist.

For presence absence data then, the Jaccard distance would be
appropriate, but as it and bray-curtis are rank order similar you can
use either. In fact, internally the same C code is used to do the
computations, the R wrapper alters the outputted bray-curtis distances
if you ask for Jaccard. If you want to bullet-proof yourself, use
Jaccard; you might save yourself from having to explain why you used
bray-curtis.

To take a stab at the first point you make, I'd argue that there isn't a
"best" ordination method. Some are better than others in some respects,
worse in other respects. You should choose a tool that you are
comfortable with, reflects the properties of your data and provides
results in an interpretable manner.

HTH

G

> 
> 
> Gavin Simpson wrote:
> > 
> > On Thu, 2009-09-03 at 04:49 -0700, swertie wrote:
> >> Hello, I read a lot about ordination, but I am still confused... I have
> >> data
> >> on species presence/absence for 8 different sites and I would like to
> >> represent my species and the sites on an ordination plot to see if some
> >> species are associated with specific sites. I used metaMDS function,
> >> which
> >> displays both sites and species and it seems to work well. However why
> >> are
> >> most people using CA instead of non metric multidimensional scaling?
> >> Thank
> >> you  
> > 
> > Inertia ;-)
> > 
> > It's what they were taught, because that's what their supervisors were
> > taught etc.
> > 
> > CA has a long history within ecological circles and most people will be
> > familiar with it.
> > 
> > nMDS is an iterative algorithm that may nor may not converge to "the"
> > solution. There may not be "one" solution but rather many equally good
> > ones. It takes a lot more effort (well, if you are outside of R) to run
> > nMDS properly and check you're not converging to a local solution.
> > 
> > There are lots of reasons. Why are interested in this; are you concerned
> > you've done something wrong or inappropriate with your data?
> > 
> > HTH
> > 
> > G
> > 
> > -- 
> > %~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%
> >  Dr. Gavin Simpson             [t] +44 (0)20 7679 0522
> >  ECRC, UCL Geography,          [f] +44 (0)20 7679 0565
> >  Pearson Building,             [e] gavin.simpsonATNOSPAMucl.ac.uk
> >  Gower Street, London          [w] http://www.ucl.ac.uk/~ucfagls/
> >  UK. WC1E 6BT.                 [w] http://www.freshwaters.org.uk
> > %~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%
> > 
> > ______________________________________________
> > 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.
> > 
> > 
> 
-- 
%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%
 Dr. Gavin Simpson             [t] +44 (0)20 7679 0522
 ECRC, UCL Geography,          [f] +44 (0)20 7679 0565
 Pearson Building,             [e] gavin.simpsonATNOSPAMucl.ac.uk
 Gower Street, London          [w] http://www.ucl.ac.uk/~ucfagls/
 UK. WC1E 6BT.                 [w] http://www.freshwaters.org.uk
%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%~%