search

igmhub / picca

set of tools for continuum fitting, correlation function calculation, cosmological fits...
GNU General Public License v3.0
29 stars 22 forks source link

p1d masking change #1064

Open Waelthus opened 3 months ago

Waelthus commented 3 months ago

change the p1d combination to reject short spectra or spectra with heavy masking based on a certain fraction of spectrum, not number of pixels.

Waelthus commented 3 months ago

The P1d tests run, but obviously results change with this change, could update the tests if we decide to merge this or let them use the previous defaults

corentinravoux commented 3 months ago

I added some comments, I do not remember the status of the tests on the number of chuncks that we can get back.

Waelthus commented 3 months ago

I think it were not many chunks, mostly because the max masked values used in current analyses are way higher than the previous defaults already... So probably not urgent to merge either. But it would allow a way to actually have the cuts defined more evenly across redshifts, the old way would allow masking significant parts of the spectrum at z=2, but not nearly as much at z=4

corentinravoux commented 3 months ago

What I suggest, and regarding the strategy in Y1: You can make how we were masking as the standard, and the masking fraction as a potential option. We can then consider the masking fraction as a variation of the Y1 baseline ? If we consider that this is a good enough improvement, we will add it to the next baseline