Closed kadrlica closed 6 years ago
My hack fix was to just start the mcmc well inside the parameter range (since the starting point is a tight ball). I have dug a little into how to fix this but I haven't implemented anything yet (on my to do list).
As an aside, a related enhancement would be to have an option to start the mcmc uniformly within the prior volume instead of a tight ball. I am interested in trying to implement this.
Both of these updates sound good to me. We are happy to accept any hacks that you implement.
@apace7 Let's see if we can fix the bug (i.e. make sure to reject points outside of the parameter space) by the end of this week. We can push the uniform distribution enhancement to another issue.
@apace7 noted that the points we are using to seed the mcmc are not explicitly constrained to the allowed parameter range. This is a bug and should be fixed. @apace7, any chance you could submit a pull request with the fix you've been using?