[R-sig-eco] specifying a "chord" distance matrix in metaMDS

[Jari Oksanen]

On 22/09/10 22:01 PM, "Mike Rennie" <mikerennie.r at gmail.com> wrote:

> Thanks Jari- however, when I try this code, it gives me some results for
> runs with stresses, says I have a solution reached, and then I get an error:
> 
> Error in cov.wt(x, x.w) : weights must be non-negative and not all zero
> 
> I'm assuming this is saying I have either zeros or negative values in my
> distance matrix? Which I shouldn't- when I specify the distance matrix
> outside of metaMDS, it's all values between 0 and root 2, and I don't see
> any zero values.

Mike,

The error message comes probably from wascores() that finds the species
scores after ordination. To be sure, you can say traceback() after getting
the error message to see how the error is propagated.

I'm sure that the computer is correct if she claims that some of your
species abundances are non-negative or that all are zero. If you think
otherwise, you're wrong. Don't argue with computers. If this is wascores()
like I assume, the error is caused by the transformed abundance data, and
not by the dissimilarities. Create transformed data with

tmp <- decostand(orig.data, "norm")

and see if there are any negative values (min(tmp), any(tmp < 0)), or all
zero columns (any(colSums(tmp) <= 0)).

Cheers, Jari Oksanen