Theta prior boundary issue

[David Pearce]

Hello,

I am trying to run migrate on a dataset that has 18 msat markers for six populations with sizes of 20, 11, 3, 28, 10 and 23. However, I keep running into an issue where the needed upper theta prior boundary is extremely high, >20,000 in some runs (full migration matrix). I have tried running migrate with only three populations and also pooling the populations into their respective side of a boundary that we are investigating. Regardless of how I try and run migrate the theta boundary is still an issue.  I look forward to hearing your recommendations. 

Best,

David

[Peter Beerli]

David,

Anything beyond an upper bound for theta > 200 for msats seem wrong, how variable is the data? If the data is not informative then the output will report the prior and thus trigger the warning that one may need to increase the upper bound.

Show me the outfile 

Peter

--
You received this message because you are subscribed to the Google Groups "migrate-support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to migrate-suppo...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/migrate-support/8cc7c4e7-d180-46b9-a865-33f15f316873n%40googlegroups.com.

[Luis Montalvo]

Thanks for answering my questions. I think the data should be fine. I ran it with the exact parameters and files you used. I am attaching the output file you got when you ran it. I sent you the data and paramfiles like one month ago and also the output I got. I always got the same. Since I am using the exact files you used and getting different results, I guess that maybe could be an issue with the installation in my cluster. I used Migrate-4.4.4. 

To view this discussion on the web visit https://groups.google.com/d/msgid/migrate-support/C8576368-55F6-454B-B791-CEDE69EDBCF4%40gmail.com.

[David Pearce]

Hi Peter,

I have attached the most recent outfile (18m) for all six populations using a full migration matrix. I would say that the data is fairly variable, even more so when I group the populations on either side of the boundary that we are looking into. I have also attached a copy of this outfile (rift).

-David

[Peter Beerli]

David,

I suggest to run the 2 population version with an upper limit for Theta of 400 and also cut back for the migration parameter to somethin like 400 or 40 because a Theta=400 and M=40 will lead to 4Nm = 1600 which seems _very_ high.

for the manypopulation run I would use an even lower upper bound than for the two populations model, but if you do a model selection with migrate

then you could actually compare which model 2-pop model vs 6-pop model. 

Your bounds are too high and this leads to histograms that peak at or near zero because the histograms has only so many bars (default is 1500).

In addition: you use a rather old version of migrate the currently supported main version is migrate 5.0.3 (but I still lack a windows binary, for that)

Peter

To view this discussion on the web visit https://groups.google.com/d/msgid/migrate-support/4260f0c1-b23f-4bd3-9ed5-58e7f19487b8n%40googlegroups.com.
<6.4.22 18m_outfile.pdf><rift_outfile_low theta.pdf>