I'd like to know if the branch to an outgroup is the same relative distance from the other OTUs,  in two different data sets.

Here roughly are the steps I followed:
I set up a simple data file with 3 otus and the two data sets concatenated.  

-I created two partitions (one for each data set) and set them to have the same tree (which is trivial for 3 otus).

- optimized the likelihood,  and save the model.

- clicked on both trees in the parameters window and selected relative ratio.  

-optimized and saved the model and clicke on LRT

With the HKY model and the transition/transv ratio applied globaly,  this gives a test with 2 DF.

However I am concerned that  the relative branch lengths being tested are all three branches in the tree.   What I'd really like to test is just the branch to the outgroup.

I tried constraining the two tip branches to be equal in length, for each locus.  I did this and could generate a likelihhood.  But when I tried to also add the relative-ratio constraint,  I was told there were no free parameters.

Thanks for any tips.

JH

Dear JH,

If you have two trees:

Tree t1 = (in1,in2,out);
Tree t2 = (in1,in2,out);

under the HKY85 model with a branch length parameter t and global tv/ts ratio of R (assume that R is the same in both data sets), then imposing a relative ratio constraint is going to be equivalent to setting

t1.in1.t := C * t2.in1.t;
t1.in2.t := C * t2.in2.t;
t1.out.t := C * t2.out.t;

where C is the proportionality parameter estimated from the data (thus the 2DF test).

You are correct that the Relative ratio test will constrain all branches (and not just the outgroup).

Unless you have a 4th sequence (or use a non-reversible model), all you can test regarding the outgroup is the amount of divergence from the MRCA of (in1,in2) - since you can't root the 'out' branch.

It sounds like what you are trying to do is test whether the ratio

t1.out.t/t1.in1.t = t2.out.t/t1.in1.t

(this assumes molecular clock on in1, in2) or something like for the general case

t1.out.t*2/(t1.in1.t+t1.in2.t) = t2.out.t*2/(t2.in1.t+t2.in2.t)

I am linking a human/chimp/mouse example with both of the above constraints defined (using the GUI).

Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login

Let me know if this helps/addresses your question.

Cheers,
Sergei