Negative blackbody radius :)

[gabrielastro]

Hello again! I am doing:

Pargrenzen = {}
Parnormprior = {}

Pargrenzen['disk_teff']   = (10., 1000.)
# needs to be present to "activate" the variable
#   even though we will set normal priors afterwards:
Pargrenzen['disk_radius'] = (0., 700.)

# eine gaußsche Kurve für 200 ± 200 RJup
Parnormprior['disk_radius'] = (200, 200.)

fit = FitModel(…, bounds=Pargrenzen, normal_prior=Parnormprior)

and get a best-fitting disc radius of… -6 RJup :).

Actually, the mean is zero because there is a peak at +6 RJ too. However, e.g. the mass posterior is not symmetric: high-mass solutions are possible with a negative blackbody radius but not with +6 RJup.

Two things here: It would be good to restrict R to be positive, of course, but I thought that if both bounds and normal_priors are set, the bounds would truncate the normal_priors. I looked again but maybe I missed it in the documentation of FitModel. Is it the same behaviour for all parameters: the final prior not being the multiplication of the top-hat from bounds and the Gaußian from normal_prior?