bamm converging too quick

[ljq...@gmail.com]

Dear colleagues,
My BAMM (2.5.0) analyses are converging well, but what intrigues me is that they converge way too fast.  In the second sample, it finds an optimum and gets stuck there. That doesn't seem right, does it?  I've tried modifying the priors, but nothing changes in this aspect. Does anyone have any insight?
Thanks!
Luiz

[pasca...@gmail.com]

Hi,

How large is your tree?

I'm not sure why you think BAMM is getting stuck... the log likelihood is varying as is the number of shifts, which is what you'd expect.

With between 0 and 2 possible shifts, I'm not surprised the analysis would get there quickly. The best way to evaluate that sort of thing is to use bayesian convergence diagnostics, like the effective sample size of the log likelihood. See here on the bamm-project website.

-Pascal

[ljq...@gmail.com]

Thank you, Pascal, for your comment.

The three is relatively large (344 tips). Given the size, I thought convergence was too fast, and that made me worry. ESS is fine. I also followed the instructions given in 'Is the BAMM likelihood computed correctly?'. Everything seems okay as well.

Thanks again!

Cheers

Luiz