Hi,

I worked on HyPhy and PAUP with same data, tree topology, and model (GTR). They gave very closed branch length for the tree but very different Log likelihood(-3594 from Hyphy, -3815 from PAUP).  

Any reasons for this different result? Thanks.

my HyPhy code:

DataSet myData = ReadDataFile Hominoid.txt");
DataSetFilter myFilter = CreateFilter (myData,1);

HarvestFrequencies (obsFreqs, myFilter, 1, 1, 1);
REVRateMatrix = [*,a,b,c}{a,*,d,e}{b,d,*,f}{c,e,f,*];
Model REV = (REVRateMatrix, obsFreqs);
Tree myTree = ((((((Ch,Bo),Gr),Hm),Or),Gb),Mc);

LikelihoodFunction theLikFun = (myFilter, myTree, obsFreqs);
Optimize (paramValues, theLikFun);
fprintf (out.txt, theLikFun);

Dear jxn,

The difference is that you have fitted 6 parameters PER BRANCH in HyPhy - i.e. a separate GTR for every single branch. PAUP assumes that those rate parameters are shared for the entire tree. Take a look at the LocalVsGlobal.bf file in Examples.

Cheers,
Sergei

Dear Sergei ,

Thank you for your reply. How can i make a global analysis on GTR model? Is V always 2?should i put V into the RateMatrix?

global V=2.0
REVRateMatrix = [*,av,bv,cv}{av,*,dv,ev}{bv,dv,*,fv}{cv,ev,fv,*];

Thanks.

jxn

Dear jxn,

Please take a look at the Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login (second part) and the files in Examples/SimpleAnalyses - especially REVshared.bf; you will find exactly what you want to know in that example.

Cheers,
Sergei

P.S. If all you are doing is fitting standard models, you may be better off using one of the standard analyses or the interface module. This is also described in the link I give above.