bug in "physignal.z()" geomorph 4.0.10

[Stylianos Koutalis]

Dear all,

Today I encountered a potential bug that did not occur last year. When running the physignal.z() function with univariate data, Csize for various Species (using the "front" lambda optimization) in the form of an [n, 1] matrix, I received the error: "Error in Y[d, ] : incorrect number of dimensions". 

This error did not occur with the same script and data in a previous analysis. The physignal() function still works fine on the same univariate input. To work around the issue, I duplicated the univariate data as a second column - new.matrix <- cbind(data, data) - and then ran physignal.z() (using "front" lambda optimization again), which reproduced my original results.

I'm reporting this as a possible bug since the behavior has changed. Thank you very much for your time and attention!

Best regards,
Stelios

[Mike Collyer]

Dear Stelios,

I believe if you update both RRPP and geomorph from Github (addresses in description files), this issue has already been addressed.  It was caused by R converting a single-column matrix into a vector, even after it was established as a matrix.  (This appears to be due to recent R updates, not the physignal.z function.)  This is why your fix worked because it forces R to retain a matrix.  We added a line of code to assure retention of a matrix, so it should work now.  Please let us know after updating if still does not work.

Best,

Mike

--
You received this message because you are subscribed to the Google Groups "geomorph R package" group.
To unsubscribe from this group and stop receiving emails from it, send an email to geomorph-r-pack...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/geomorph-r-package/8971e4d5-6775-46c5-8824-41e3c2f6f30dn%40googlegroups.com.

[Stylianos Koutalis]

  Dear Professor, 

Good morning. I had the same problem again, this time with the overall result of the physignal.z function. The time we discussed the issue with the bug, which indeed was solved when i  updated (or manually go around) RRPP,  now, the same dataset, results in this :

> PS.shape.z <- physignal.z(A = extant, phy = tree.all, iter = 10000, lambda = "front") 

> print(PS.shape.z) 

Call: physignal.z(A = extant, phy = tree.all, lambda = "front", iter = 10000) 

Evaluation of phylogenetic signal effect size 

Optimization method: front 

Optimization performed in 11 data dimensions. 

The scaling parameter, lambda, was optimized at 0. 

This means that the log-likelihood was invariant across permutations 

and there is no phylogenetic signal in the data.

The same dataset, when the 'front' optimization is removed, works as expected. I mean, this works now as well, but I'm unable to recreate my original result.

Thank you for your time! Looking forward to hearing from you.

Best,

Stelios