some questions regarding EBSP

[Ramiro Morales-Hojas]

Dear Beast users,

I am new to extended bayesian skyline plots and I would like to ask some questions that could be very basic but that I would like to understand. I am running an analysis with 16 nuclear loci, and I have an estimate of 0.004 substitutions per million year for the species which I have included in the analysis as a clock rate following the tutorial. I have run three times the analysis with 10 (log 1000), 100 (log every 10000), and 200 million generations (log every 20000).

My first question relates to the low ESS of the posterior and prior I get, indeed red numbers (the rest of the parameters have quite high ESS values), and that don't increase significantly from the first preliminary 10 million run, to the 200 million chain run, although I have seen somewhere in this group that the ESS values don't necessarily have to be high in skyline analyses. Nevertheless, I wanted to combine the 100 and 200 million analyses to see if it improved. Using Logcombiner (with 10% burnin for each lo file) I get a file that has 180 million states shown in Tracer instead of the 270 million states I was expecting (200mill-10% + 100mill-10% should give 270 million states, is this correct?). It doesn't matter which version of LogCombiner I use, I never get the expected number of states. Obviously I'm doing something wrong here, any idea of what is going on?

My second question relates to the clock rates I get for the genes after the analysis. Having introduced the 0.004 subst/million years as clock rate for one all genes (although I still alowed the analysis to estimate all rates except for the first gene) I get estimates for the other genes that range from 0.1 to 0.8. I'm not sure how to interpret these rates, aren't these estimates too high to be substitution rates? Do they refer to time in relation to present? Any idea on how can interpret these or if they appear to be wrong? All loci I'm using are genes from two autosomes and I expect similar rates. Maybe I should fix this rate for all genes unchecking the "estimate" box for all.

The third question I have is how do I interpret the population size change, which I get a value of 2.056 (very stable in the different runs and high ESS value). Does it mean that there have been two changes in the population size? When I plot the csv table I get a plot that looks like having just one single increase.

My last question is more practical. I have plotted the csv table in Excel to obtain the extended bayesian skyline plot but, is there a way to get the skyline plot using R?

Many thanks and I hope that my questions make sense.

Ramiro