I've seen a few instances where the calculated hessian for the MLE fits gives invalid results. I only saw this when quite a few params were floated. I suspect (but haven't dug into it) that the underlying issue here is that the minimizer was failing to converge properly. If this pops up again, we should investigate and fix (or, at least make sure to fail gracefully).
I've seen a few instances where the calculated hessian for the MLE fits gives invalid results. I only saw this when quite a few params were floated. I suspect (but haven't dug into it) that the underlying issue here is that the minimizer was failing to converge properly. If this pops up again, we should investigate and fix (or, at least make sure to fail gracefully).