0

I used the envfit() in the vegan package to determine the environmental gradients controlling the 2 NMDS axes. I had 2 categorical environmental (factors) and 4 continuous variables (vectors).

My question is does vector fitting give degrees of freedom? If yes, how and where do I find it? If no, why not?

tg110
  • 401
  • 1
  • 3
  • 15

1 Answers1

0

It does not as envfit is a purely correlative approach. You can test as many variables as you want with envfit, the ordination won't change and the fit of the other variables won't change.

Richard Telford
  • 9,558
  • 6
  • 38
  • 51
  • then is the r2 in the results a correlation coefficient, if so, which one? or is it the usual R square as in explanatory power/goodness of fit? – tg110 Apr 26 '16 at 22:16
  • I assume you understand that it is both: it is the square of correlation coefficient **and** the usual measure of "goodness of fit". There is an arrow, you project your points to that arrow and calculate a correlation between projected positions and the variable. To see how this works, use `ordisurf` with argument `knots = 1`. – Jari Oksanen Apr 29 '16 at 12:33