Agustin Lobo wrote:
> I disagree (in part). Although it is true that the stat. analysis is
> dependent upon scientific conceptions, it is not true that clustering
> species using their abundances in relevees as descriptors necessarily
> implies a "sigmatist" conception. Clustering, in this context, is much
> more "neutral" than that: just finds sets of species that tend to be
> found together in the same relevees. It is the explanation to this
> observation which implies a conception of nature. In my own
> point of view, this is just the consequence of (i) several gradients
> introducing each their own ordination, (ii) second-order interactions
> among species introducing complexity and non-linearity. These
> 2 reasons promote that the co-occurrence of some species be
> more likely than expected under an independent and linear
> ordering of each species to each gradient.
>
> I also would like to add that a "flexible" classification
> goes beyond the concept of fuzzy classification. A flexible classification
> can be achieved using maximum likelihood but considering the
> vector of likelihoods as well as the maximum one. That is,
> we can define few, clear typical communities, and understand
> that most individual relevees will be somehow
> intermediate, according to their likelihood to each of the
> models.
I completely agree with this point of view!!!
If you are interested with such flexible classification mixed with ordination, I suggest you to try out Mulva-5 software, which you can download at this address:
http://www.wsl.ch/land/MULVA/M5load.html
I organise with O. Wildi and others a postgraduate seminar and practical course (free of charge!) about "Statistical & dynamic approaches to ecological modelling", september 15-17, 1999 in Neuchâtel. Please contact me if you are interested!
François
--
* * * * * * * * * * * * * * * * * * * * * * * * *
* Francois Gillet
* Maître-assistant
* Laboratoire d'écologie végétale
* Institut de Botanique
* Université de Neuchâtel
* Emile-Argand 11
* CH-2007 NEUCHATEL
* Switzerland
*
* E-mail: francois.gillet@bota.unine.ch
* URL: http://www.unine.ch/biol/levp/fg.html
* Tel: +41 32 718 23 17
* Fax: +41 32 718 21 01
* * * * * * * * * * * * * * * * * * * * * * * * *
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:36:00 MET