I think there might be an issue with how the uncertainties are being plotted in the regularised example, as sometimes, the single observation uncertainties are less than the regularised. This is probably due to trying to get around numerical issues in the estimation of the inverse Hessian. The code that deals with this is quite messy, so likely that I made a stupid mistake.
I think there might be an issue with how the uncertainties are being plotted in the regularised example, as sometimes, the single observation uncertainties are less than the regularised. This is probably due to trying to get around numerical issues in the estimation of the inverse Hessian. The code that deals with this is quite messy, so likely that I made a stupid mistake.