Internal error in TBR calculation using landmark data

[Pedro H. Costa]

Hi everyone,

I am having trouble running heuristic searches with NT algorithms using a matrix with landmark characters.

My searches keep crashing at random moments with the message "internal error in TBR calculation", which halts the entire process.

I found that removing certain NT options (such as tree fusing) and including the command lmark errmarg 0.00001 in my .run script made the analyses run for a significantly longer time. However, I don't know if altering this error margin is recommended, and the search still eventually crashes with the same internal error.

Could anyone shed some light on how to bypass this problem (even in the middle of the World Cup)?

Abraço, Pedro Costa

  (P.S.: My matrix has 91 taxa and 57 characters: 6 2D landmarks, 22 continuous linear, and the rest discrete. I am running TNT v. 1.6 via command line on my University's Linux cluster)  

[Edward SIddall]

With 91 taxa and only 57 characters I don’t think tree fusing is giving you much lift.

Another option is to do multiple more conventional searches, maybe SPR if TBR is crashing out, saving the trees, and then try fusing on them afterwards.

--
Has recibido este mensaje porque estás suscrito al grupo "TNT-Tree Analysis using New Technology" de Grupos de Google.
Para cancelar la suscripción a este grupo y dejar de recibir sus mensajes, envía un correo electrónico a tnt-tree-analysis-using-n...@googlegroups.com.
Para ver este debate, visita https://groups.google.com/d/msgid/tnt-tree-analysis-using-new-technology/62f6baa1-f2ac-475a-b31c-55f6434d5eebn%40googlegroups.com.

[Edward SIddall]

But with half as many characters as taxa your problem is under defined anyway. 

[George D. F. Wilson]

"Under defined" meaning for a fully resolved tree, the number of characters in this case should be in the range of 90 - 181. In a perfect binary tree where every character defines a branch, the number of characters is 181, which is a "rule of thumb" if building a new dataset. Homoplasy  increases the number of characters necessary so more are needed. I have reviewed more than one manuscript recently where the authors produced a result where the number of features/characters was too low for even a usable majority rule consensus tree to be found. {re: previous, dislexia strikes again...}

George D.F. (Buz) Wilson
Saugatuck Natural History Laboratory
email: buzw...@gmail.com

Para ver este debate, visita https://groups.google.com/d/msgid/tnt-tree-analysis-using-new-technology/D7B6E527-43B6-4FDD-949F-7AB614B724AF%40gmail.com.

[Pedro H. Costa]

Hi everyone,

I just realized I made a mistake and swapped the dimensions: my matrix actually has 91 characters and 57 taxa (not the other way around). Maybe my dyslexia is to blame too, haha.

Anyway, thank you for the replies and explanations! I'll follow the tip and run the searches using SPR instead of TBR.

If the runs work out, I'll share the results here.

Abraço,