Hi everyone,

Our group has been discussing Mantel's tests and the appropriate permutation tests to assess significance. We would like to find out if there is a consensus on which software packages are best - i.e. do the correct permutation tests (permute residuals). Any suggestions would be greatly appreciated.

Thanks.

Daria

Some packages we know of that perform Mantel's tests:

R package

Vegan for R

FSTAT

Arlequin

The problem is that most programs do not explicitly state how they do the randomization.  Also, there are actually several ways for permuting "residuals" for the partial Mantel test and I think the permutation of residuals of a null model (Anderson & Legendre 1999) is appropriate, while the permutation of one residualized matrix (Smouse et al 1986) is not.  Another (safe and in my view most intuitive) option is to simply permute the "dependent" matrix...and it's what's implemented in R package ecodist.  There is still a lot of debate with respect to the most appropriate permutation method for the partial Mantel test...

If the software you want to use does not explicitly state which method is used, you might want to check with the author of the program.

Hope this helps and doesn't add to the confusion :O)

Niko