Further to my notes on the BEAST template SNAPP from a few days ago, here is a little issue that people who happen upon this blog via entering SNAPP into a search engine might be interested in:
Check the 'totalcount' values in your XML files before starting the BEAST run.
You would think that if you entered a SNP dataset with the three character states 0, 1 and 2* into BEAUTi, it would recognise three character states and write totalcount="3" into the XML file, wouldn't you?
Well, I thought so too, but apparently it will, at least as version 2.1.3 on my machines, randomly pick either a 2 or a 4. No idea why. Nor do I have any idea how BEAST interprets the 0, 1, 2 states when it only expects two states. What I do know is that you can correct the XML file manually - simply do a search and replace of totalcount="(whatever the wrong number is)" with totalcount="3" -, and that my analyses are even slower now that I have corrected BEAUTi's mistakes. But at least this explains why I got nonsensical results in some cases. Yippee.
*) Homozygous for the major allele, heterozygous, and homozygous for the minor allele, respectively.