search

brettc / partitionfinder

PartitionFinder discovers optimal partitioning schemes for DNA sequences.
Other
61 stars 44 forks source link

Add support for additional RAxML models #23

Closed roblanf closed 9 years ago

roblanf commented 9 years ago

RAxML 8.x has a lot more models available. Easy to add support for these.

roblanf commented 9 years ago

Note this from the latest RAxML manual too:

"By appending X to the model strings you indicate that you want to use a maximum likelihood estimate for the base frequencies"

We could easily add this, e.g.

GTR, GTRX, GTR+G, GTRX+G, etc.

We could use the same notation for PhyML models too, which have the same options (except you specify the models through the commandline)

cmayer commented 9 years ago

Hi,

I do not understand this. The GTR model does include estimates for base frequencies inherently. How could this be switched on or of. Its a bit like B&B (Bed and Breakfast) with or without a bed.

Best Christoph

Am 18.01.2015 um 03:46 schrieb roblanf:

Note this from the latest RAxML manual too:

"By appending X to the model strings you indicate that you want to use a maximum likelihood estimate for the base frequencies"

We could easily add this, e.g.

GTR, GTRX, GTR+G, GTRX+G, etc.

We could use the same notation for PhyML models too, which have the same options (except you specify the models through the commandline)

— Reply to this email directly or view it on GitHub.

Dr. Christoph Mayer Email: c.mayer.zfmk@uni-bonn.de Tel.: +49 (0)228 9122 403

Zoologisches Forschungsmuseum Alexander Koenig

Stiftung des öffentlichen Rechts; Direktor: Prof. J. W. Wägele Sitz: Bonn

roblanf commented 9 years ago

The difference is that they can be estimated empirically (just add up all the bases in the alignment) or via ML. If they are estimated via ML that costs you 4 extra parameters. Usually empirical estimates are fine.

On 19 January 2015 at 04:14, cmayer notifications@github.com wrote:

Hi,

I do not understand this. The GTR model does include estimates for base frequencies inherently. How could this be switched on or of. Its a bit like B&B (Bed and Breakfast) with or without a bed.

Best Christoph

Am 18.01.2015 um 03:46 schrieb roblanf:

Note this from the latest RAxML manual too:

"By appending X to the model strings you indicate that you want to use a maximum likelihood estimate for the base frequencies"

We could easily add this, e.g.

GTR, GTRX, GTR+G, GTRX+G, etc.

We could use the same notation for PhyML models too, which have the same options (except you specify the models through the commandline)

— Reply to this email directly or view it on GitHub.

Dr. Christoph Mayer Email: c.mayer.zfmk@uni-bonn.de Tel.: +49 (0)228 9122 403

Zoologisches Forschungsmuseum Alexander Koenig

  • Leibniz Institut für Biodiversität der Tiere - Adenauerallee 160 53113 Bonn, Germany www.zfmk.de

Stiftung des öffentlichen Rechts; Direktor: Prof. J. W. Wägele Sitz: Bonn

— Reply to this email directly or view it on GitHub https://github.com/brettc/partitionfinder/issues/23#issuecomment-70416382 .

Rob Lanfear School of Biological Sciences, Macquarie University, Sydney

phone: +61 (0)2 9850 8204

www.robertlanfear.com

roblanf commented 9 years ago

to clarify (and so I don't forget), ML base frequencies cost you 4 parameters, but only 3 free parameters, because they sum to 1.0

roblanf commented 9 years ago

Done. The default is to use empirical base frequencies, but all of the models are now in there. Including all the protein models, AND a GTR protein model.

See models.csv