metaMDS(..., engine ="isoMDS") produces error

[martinmodrak]

I have frequent convergence problems with the default setting of metaMDS so I wanted to try isoMDS instead of the default, but the code does not complete. Reproducible example follows:

library(vegan)
library(MASS)
data(varespec)
metaMDS(varespec)  #Completes, but does not converge
metaMDS(varespec, engine = "isoMDS")  #Does not complete

The error from the last line is

Error in out$dist/scl : non-numeric argument to binary operator

I am using vegan 2.5-4 with R 3.5.3 on Windows

[jarioksa]

Yes, this is a bug: isoMDS does not return information you need for this scaling. However, you really should not use isoMDS – it has several problems. However, I'll fix this.

[martinmodrak]

1. Thanks for the quick reaction. 2) I understand there are issues with isoMDS, I was mostly just curious what would happen 3) Sorry if I sounded like demanding support, I definitely don't need this to be fixed to carry on with life.

Thanks for the nice package!

[jarioksa]

Don't worry: we think that if we provide an option like engine the option should work! I am not offended or disturbed of reports of failure, but I'll do my best to fix the problems, and I'll fix this one. We just did not have a test to the isoMDS engine and this fault slipped through when we made an unrelated change.

About your actual problem of no-convergence: Have you read the section "Convergence Problems" in the metaMDS help (?metaMDS), and followed those procedures?

[jarioksa]

Fixed with commit 8ad3e80