Low ESS Mutation rate and Clock rate

[Ian AB Petersen]

Hello,

I'm running a beast analysis with ~60 taxa and 5 gene partitions(3 mito 2 nuclear) with a single fossil calibration point 20mya. I've been having trouble with low ESS in my analyses. I've tried many ideas I found here (beast*, fix mean subs rates etc) and was able to increase ESS in almost all areas by deparameterizing my models (HKY instead of GTR) and by using a strict clock rather than relaxed log normal. However, I still have very low ESS for mutation rate and clock rate across gene partitions. Any advice would be awesome! Attached is the file if it's useful.

Thanks,

Ian Petersen 

[Remco Bouckaert]

Hi Ian,

The XML you posted shows that both relative substitution rate and clock rate are estimated. The total rate on a branch is the product of these two (for strict clock), so a clock rate can be increased for one partition and as long as the relative substitution rate decreases proportionally another set of substitution rates and clock rates can be found resulting in the same branch rates. In other words, substitution and clock rates when estimated for each partition individually are unidentifiable, resulting in the low ESSs you observed.

To get better ESSs and have individual rates for each of your partitions fix all substitution rates to 1, and do not estimate them (uncheck the estimate box in BEAUti next to substitution rate in the site model panel). 

Cheers,

Remco

--
You received this message because you are subscribed to the Google Groups "beast-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to beast-users...@googlegroups.com.
To post to this group, send email to beast...@googlegroups.com.
Visit this group at http://groups.google.com/group/beast-users.
For more options, visit https://groups.google.com/d/optout.
<beast_100m_HKY_Strict_FixMeanSubRate.xml>