[R-sig-eco] Multicollinearity of environmental variables in unconstrained ordination

[Ralf Schäfer]

Dear list,

I have a question regarding how to deal with multicollinearity of variables that enter the Envfit procedure in NMDS {vegan}.
One option is just to do nothing and only mention the intercorrelation between variables (compare et al Salemaa (2008) Boreal Environment Research).
Other options I know of is to omit the less relevant variable from the Envfit or only use the residuals of this variable (as suggested by Graham (2003) Ecology, 84, 2809-2815.). Finally it is possible to do a PCA on the variables before.

Anyway, are there better solutions to this problem how to deal with multicollinearity of environmental variables that are used for fitting vectors (or surfaces) over a NMDS ordination?
If so, I would appreciate if you could also outline briefly how to implement this solution in R if possible.

Cheers
Ralf Schäfer

------------------------------------------------------------

Prof. Dr. rer. nat. Ralf Bernhard Schäfer
Juniorprofessor for Quantitative Landscape Ecology
Environmental Scientist (M.Sc.)
Institute for Environmental Sciences
University Koblenz-Landau
Forststrasse 7
76829 Landau
Germany
Mail: schaefer-ralf at uni-landau.de
Phone: ++49 (0) 6341 280-31536
Web: http://tinyurl.com/37ot8xq