DynamicPSF Moffat fitting

[mschuster]

I would have thought that fit produced by a general Moffat function with an arbitrary beta parameter would be as least as good as the fit of a Moffat function with a fixed beta, but this does not seem to be the case, as the attached image shows. For the first star, the Moffat4 MAD is less than the MAD of the general Moffat. For the second star the Moffat15 and Lorentzian are better. Is this expected behavior?

[Juan Conejero]

Hi,

In general, yes. In the first case the difference is insignificant (1.324x10^-2 and 1.325x10^-2). In the second case the difference with Moffat15 also has little significance. The difference with a Lorentzian fit is somewhat larger for the second star (although 2.7x10^-3 is not really large), but the fitted beta parameter is 1.8, which is (morphologically) not far from beta=1.

All of these results and variations are perfectly normal and due to the inherent uncertainty of the PSF fitting process. However, I'll think on a more robust quality estimator for a future version of the DynamicPSF tool. Thanks for questioning.