underflow on large trees about transphylo HOT 1 CLOSED

Hi Gerry,

Thanks very much for reporting and solving this problem! There was only a little problem with your code in the case where the r parameter of the NegBin offspring distribution is close to zero, in which case there is an exact solution (cf lines 73 and 106 of src/probTTree.cpp). I've corrected this and committed the changes.

Yes, you're right that the procedure to optimise the MCMC starting point can sometimes try a bit too hard. You can of course start from a non-optimised point, but you might then find that it takes a long time for the MCMC to converge. Alternatively you could reduce the number of attempts at optimising on line 53 of R/makeCTreeFromPTree.R from 100 to 10 or even 2 or 1.

Best wishes,
Xavier

from transphylo.