[R-sig-eco] ANOSIM

[Jari Oksanen]

On 23/11/2012, at 18:44 PM, <aristide.andrianarimisa at blueline.mg>
 <aristide.andrianarimisa at blueline.mg> wrote:

> Dear responsible,
> 
> I want to know if ANOSIM (Global R, and p-value) are only use for
> dissimilarity matrix as well as similarity matrix
> 
> If I use similarity matrix like Jaccard, is still the same interpretation
> of the output R= completely dissimilar and R = 0 completely similar?
> 

Aristide,

If your are using the vegan implementation of anosim, the answer is: no, you cannot use anosim() for similarities but only for dissimilarities.

If you refer to the *method* (instead of *implementation*), the answer is: yes, of course you can use ANOSIM for similarities. You only need to implement it that way (change the signs, reverse the P-values compared to vegan::anosim()). 

The vegan implementation is made for R, and R deals with dissimilarities. Some packages may change this R convention, but similarities are not R compliant.

Cheers, Jari Oksanen
-- 
Jari Oksanen, Dept Biology, Univ Oulu, 90014 Finland