Phycas woes

I have been having very good luck with using Phycas to run stepping stone analyses to estimate the marginal likelihood of different partitioning schemes. I've obtained the marginal likelihood estimates of all of the partitioning schemes I intend to try for one of the phylogenies I am building...except for the last one.

This last one happens to be the most parameter-rich scheme, with every gene partitioned by codon position. Perhaps this is why it is having issues. The stepping stone MCMC ran smoothly, although it took over four days to run. The issue happened at the very end as it was calculating autocorrelations and estimated sample sizes:

Since what I really care about is the marginal likelihood estimate, I wanted to see if I could bypass the autocorrelation calculations and go straight to that. My lack of knowledge about Python was fairly limiting in this respect, but I managed to comment out the appropriate lines from the SumPImpl.py script within the Phycas package to only run the marginalized likehood calculations. Unfortunately, this is what I see then:

Perhaps the fact that the autocorrelation for the log likelihood was NA for beta = 0.0 has something to do with this... I think I will eventually try to run this script again when I have no others running to see if a different MCMC run will work.