darioflute / fifipy Goto Github PK

Hi Dario,
After implementing cubik for updated uncertainties for our three galaxies and refitting the cubes to get line intensities, we are getting drastically decreased signal-to-noise in the derived line intensities by about a factor of 10 for each galaxy. This is leading us to believe that cubik is overestimating the uncertainties.

I have looked through the cubik code, and there are not issue with the implementation of the weighted mean and its error. However, I noticed that you are then dividing the variance of the weighted mean by an area factor (the number of pixels under the FWHM of the region on the map). Yet, in your comment on the weighted mean you state: “the flux is already renormalized to the area of the pixel. So, the computed error is already correctly normalized to the area of the cube pixel.” So, is this division by the area factor necessary? If not, it would explain the overestimated uncertainties as the square root of the inverse of this area factor is increasing the uncertainty by approximately a factor of 2.5, at a radius/HWHM = 1. This overestimate becomes even larger as the radius/HWHM increases, which is the case for our galaxies. For example, we are using radius/HWHM = 1.35 as to allow for all pixels to have recalculated uncertainties, which results in a potential overestimate of the uncertainties by a factor of about 3.5.
Besides this, I could not find any issues with the weighted mean and its error estimates that are outputted from cubik. If you could let us know about this, we would greatly appreciate it!

Thanks,
Keith(Doore)