Jackknife problem

[Buz Wilson]

Hello POY group

I'm using POY 4.1.3 Development build on an 8 processor debian-based
machine with the flat interface & mpi, so I'm working entirely from
scripts. I haven't been able to get jackknife supports other than 1.0
for all nodes for tcm other than the default 1,2. If I run the example
script using 28S.fas, I get reasonable values for the nodes (some less
than 1.0), but if I use
<code>
read ("Ro_partitioned.fasta", "Ro_final.tre")
transform(tcm:(2,1), gap_opening:1) (* tcm used in Wilson 2009 *)
transform(static_approx)
calculate_support (jackknife:(remove:0.50, resample:1400))
report (supports:jackknife)
</code>
All of the nodes are 1.0, even if I put the % removed up to something
silly like .95. Bootstrap shows the same behaviour, although I haven't
tested that thoroughly.

Is this a bug, or is this script set up incorrectly?

cheers

Buz

[Ward C Wheeler]

Buzz--sorry so long to reply.

Are there multiple frags in Ro_partitioned.fasta? POY jacknife samples on

characters--i partitioned files, each partition is a character.

W

--
You received this message because you are subscribed to the Google Groups "POY - Phylogenetic Analysis Software" group.
To post to this group, send email to po...@googlegroups.com.
To unsubscribe from this group, send email to poy4+uns...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/poy4?hl=en.

Ward Wheeler

Division of Invertebrate Zoology

American Museum of Natural History

Central Park West at 79th Street

New York, NY 10024-5192

USA

whe...@amnh.org

http://research.amnh.org/scicomp/ward_wheeler.html

http://research.amnh.org/scicomp/projects/poy.php

[Buz Wilson]

Hmmm. I provided a topology and got a static approximation. In this
case, each site should be treated as a character, or so I thought. I'll
look at this after I get back to Oz (at the Field Museum now).
Thanks for the advice.

Buz

[Ward Wheeler]

Buz--when using a static approx for jackknife (or any other support for that matter), the homologies are conditioned on the tree (indels have no homoplasy for instance), hence resampling support is inflated relative to what you are likely expecting.  Bremer on DO for a sequence might be a better measure for you.  If you have several loci (or partitions) that can be sampled (as in EST data)  then the numbers may make more sense.

W

--
You received this message because you are subscribed to the Google Groups "POY - Phylogenetic Analysis Software" group.
To post to this group, send email to po...@googlegroups.com.

To unsubscribe from this group, send email to poy4+unsubscribe@googlegroups.com.

For more options, visit this group at http://groups.google.com/group/poy4?hl=en.

--