Retrieving canonical coefficients and correlation coefficients according to ter Braak using vegan package cca() function
Joris,
On 14/06/2012, at 17:39 PM, Joris Meys wrote:
Hi all, I get results that differ from the table given by ter Braak. I can explain the differences in sign (that's normal, as the solution of cca is defined up to the sign), and I can believe there would be small differences as we're talking different ways of estimation, but the canonical coefficients for the second axis differ substantially. My question : - Did I use the correct methods to extract what ter Braak calls "canonical coefficients" and "correlation coefficients"? - is there a logical explanation for the big difference in some results ?
I haven't checked that paper, and can't do that for several days. However, one essential difference in "canonical" coefficients between Canoco and vegan::cca is that Canoco standardizes all constraining variables to unit variance whereas vegan uses unstandardized variables. Does it help if you use Model <- cca(hs.spec ~ ., data = scale(hs.ev)) ? Cheers, Jari Oksanen
Jari Oksanen, Dept Biology, Univ Oulu, 90014 Finland jari.oksanen at oulu.fi, Ph. +358 400 408593, http://cc.oulu.fi/~jarioksa