Closed SpaceyCaseyLynn closed 5 years ago
The problem here is that the logk
parameters are running away to -infinity and essentially causing a floating point overflow. This can happen when working in the logk
basis with poor constraints on k. The prior overwhelms the data and drives the fit to K=0. Try imposing priors on one or both of the logk
parameters or fit using the per tc secosw sesinw k
basis.
Thanks BJ!
Hi, I have the attached config file, and I keep getting the error lnprob returned NaN when I try to run radvel mcmc -s file. I've tried troubleshooting for a while (+ with my advisor Lauren Weiss) and we can't figure out why. If someone can help out I'd be very grateful. k102.py.zip