Validate model for shape calibration use-case

[EiffL]

One possible application of these models is for the calibration of shape measurement methods. In order to test that the models are of high enough quality for this purpose, I guess the best strategy would be to run a simulation campaign use mock and real galaxies, use re-gaussianization on both sets of images and check that the biases are consistent.

[rmandelb]

That would show they are sufficient for this purpose, but not that they are necessary. You'd have to show that reasonable simplifications like sersic profiles don't get the biases right... which might be hard since model bias for regauss is not very large.