uncertainty estimation seems wrong

[cylammarco]

It is reported that the error estimations are wrong: empty when directly using minimiser, way underestimated when using MCMC.

[cylammarco]

When choosing the minimizer, the minimize method will not have the uncertainties computed. With the least-squares method, the uncertainties will be calculated using the jacobian (see https://stackoverflow.com/questions/42388139/how-to-compute-standard-deviation-errors-with-scipy-optimize-least-squaresif you), we opt for a more computationally heavy method to allow solutions even when the Jacobian is close to degenerate. The uncertainties are not renormalised by the reduced chi^2, however, we have the chi2 and the dof computed so that the uncertainties can be renormalised post-hoc. When using emcee the 16th and 84th percentiles of the sample are reported, the mean deviation from the 50th percentile is reported as the 1s.d.