argiopetech / base

Bayesian Analysis for Stellar Evolution
http://webfac.db.erau.edu/~vonhippt/base9/
11 stars 4 forks source link

check quality of burnin before continuing #15

Closed tedvh closed 5 years ago

tedvh commented 11 years ago

If the photometric errors are overly small on an otherwise decent dataset, or if there are other problems, the burnin can fail. When this happens it typically has logPost = -inf for every iteration and/or it won't sample (move around parameter space).

It would be nice to add a feature to check whether either of these things had gone wrong during the burnin, and if so, avoid the matrix inversion and the subsequent code failure, and instead exit gracefully with a useful message.

Low priority.