[Brendan Halpin]

I've never seen TWED produce non-metric distances. Is it your substitution matrix?

[Tom Kemeny]

I'm using exactly the same substitution matrix that I used from OM. As above, you helped me to fix the problems of non-metricness there. But the same substitution matrix fails the triangle inequality test for twed.

[Brendan Halpin]

The algorithm for TWED is not supposed to generate non-metric dissimilarities, but that is not to say that there are not bugs in the code. If you wouldn't mind sending me some of your sequence data, and the full syntax you used to run it (especially defining the substitution matrix), I would like to test it. Just the sequences mentioned in the output of metricp would be enough, and just the sequence data:

keep st1-st28 lth
keep if inlist(_n,1,2,213,7,11,56,....)

It might be better to send it to me by e-mail, at [email protected], than to post it.