Mean rate in Tracer

[Mark]

Hi,
Alexei I have contacted you previously. I have a number of radiocarbon
dated samples which I have inputted to BEAUTI as years BP, and I am
using these to estimate mt control region mutation rate in a non-human
mammal. The mean rate I am getting is 4.91E-6. What does this value
represent? s.s.yr, or for the entire sequence, or something else? It
seems extremely fast. Sorry for the naive question. Thanks, Mark

[alexei....@gmail.com]

The units are substitutions per site per year. Here are a couple of
things to consider:

(1) relaxed clocks should generally not be used on population level
data - they are generally not necessary and cause issues with
overparameterization. If you aren't already you should be using a
strict clock model.

(2) Have you checked that there are no priors that are inadvertantly
affecting your results? BEAST sets up a number of uniform priors by
default and the upper limits on these priors can often be very
inappropriate for a particular analysis. You should always make sure
that your posterior distribution isn't butting up against any of the
upper limits in the prior. If it is this could be having a strong
affect on the mean of other parameters in your analysis. An
inappropriate upper limit on population size is often a culprit for
apparently overestimated mutation rates, since these two parameters
together describe the overall level of genetic diversity.

(3) How much diversity is in your sequences? If you have very short
time frames ( <10,000 years) or very short sequences (<500) or small
numbers of sequences (<30) or a large population size compared to
sampling interval, then estimation becomes hard (see Drummond et al
2003 for a discussion of estimation power as a function of population
size and mutation rate * sampling interval). If estimation becomes
hard then the estimate of mutation rate can often be biased one way or
the other. You should look at the 95% credible interval and see
whether you believe its lower bound or not... We are in the process of
doing a simulation study to get a more systematic understanding of the
region of parameter space that is most susceptible to this problem.