Why are the uncertainties quoted in “fit.log\ sometimes so small?
• From a purely statistical (i.e. random errors) standpoint, the uncertainties quoted by GALFIT are most likely to be o.k.. Purely idealized simulations show that the values quoted by GALFIT correspond to the true definition of uncertainty. This has to be the case given that, in idealized situations, the errorbars and the Hessian matrix needed to calculate the image gradient for GALFIT to converge are intimately related. If the errorbars are wrong, the convergence on a global solution would be hampered. The key point here is that for the uncertainties to have any meaning, real galaxies have to have the profiles that we fit to them. When this is true, any profile deviation between the model and the data comes strictly from Poisson fluctuations alone. More often than not, the situation just described is a bad assumption for real galaxies because they have structures that can’t be modeled. The more inappropriate is the function being fitted for a galaxy, the *smaller* would be the statist
