Well, I hope to help the discussion on Detrended Analysis sending this
Bookmark:
http://www.helsinki.fi/~jhoksane/index.html
There is some intresting software there, and a page on CANOCO and DECORANA
bugs.
Hope that help.
Federico Spinazzi
Well, I hope not confuse the discussion on Detrended Analysis sending this
text:
P.S.: For me the fact that (apart from bugs) the analyses carried on with
ADE are repeatible is a very good thing. I think that you don't need
advanced tecniques (DCA, RA, etc.) to explore your data (pheraps I'm
wrong). These tecniques, I think, are more suitable for 'modeling'
pourposes, expecially if the specie response to the enviromental variables
is unimodal.
Does implicitly COA|Correspondence Analysis perform
an ordination on this model ? ter Braak CJF seems to suggest such a
rationale for Corresp. Anal. Why, then, the arch effect ? Because the
specie response is unimodal...?!?!
If we can see the arch effect we know that
probably the response is unimodal. So no longer Corresp. Anal. allows an
optimal specie/site ordination. I think that ter Braak CJF worked around
Corresp. Anal. in order to give a means that allows to estimate an
enviromental variable from data on ecological comunities, he didn't so for
exploratory pourposes. Am I right ?
If the question is that, we can continue to use CA as an exploratory
tecnique and we can continue to use other tecniques to extimate optimum
for species (also for making calibration studies)
I'm asking: why we don't use an algorithm (well, a program) that performs
a gaussian regression or a gaussian logit regression instead of detrended
tecniques ?
My ideas about ter Braak CJF works, papers and books are very confused ...
regards.
This archive was generated by hypermail 2b30 : Sat Feb 10 2001 - 10:21:40 MET