[R-sig-eco] dissimilarity and species turnover

[Etienne Laliberté]

Robert,

Jari explained well the main mathematical problem with the method, which
is convincing enough in itself. Reading the Legendre and others vs
Tuomisto discussion (as Jari pointed out) would be useful (the whole
thing is available as a single PDF file from
http://www.elaliberte.info/Discussion_analyzing_beta_diversity.pdf), but
to make things short, consider also this additional point.

Your initial question (i.e. whether productivity affects species
turnover) could be directly tested using the "raw-data" approach (e.g.
through RDA after species transformations, or maybe CCA) - there is
simply no need to transform your response and explanatory variables into
distance matrices to test your hypothesis in the "distance world" (.
Legendre et al (2005, Ecol Monogr) had shown that doing so (on top of
being mathematically doubtful) leads to less powerful tests (i.e.
greater type-II error).

In conclusion, do not use it because of what Jari said. But to convince
you further, even if you were willing to ignore the non-additivity
problem (which you shouldn't), you'd be using less powerful tests
anyway. Bottom line: I'd say use RDA or CCA for your question.

-- 
Etienne Laliberté
================================
School of Forestry
University of Canterbury
Private Bag 4800
Christchurch 8140, New Zealand
Phone: +64 3 366 7001 ext. 8365
Fax: +64 3 364 2124
www.elaliberte.info