search

moustakas / legacyhalos

Code and papers for a multi-faceted study of the baryonic content of dark matter halos in Legacy Survey imaging.
MIT License
4 stars 7 forks source link

handle (small) negative values in WISE PSF models #141

Open moustakas opened 3 years ago

moustakas commented 3 years ago

From me:

Is it expected that 40%-50% of the WISE PSF model pixels would have negative values (at least in the 
tile I've been working on)? Should I clamp these pixels to zero? The negative values are very small after 
normalizing the PSF patch to unity, so the effect on photometry is surely negligible.

import numpy as np
from unwise_psf import unwise_psf
for band in [1, 2, 3, 4]:
    psf = unwise_psf.get_unwise_psf(band, '2325p424')
    psf /= psf.sum()
    neg = psf < 0
    print(np.min(psf[neg]), np.max(psf[neg]), np.median(psf[neg]), np.sum(neg)/psf.shape[0]**2)

-6.678769e-08 -5.6388393e-14 -3.4602092e-08 0.4885585798816568
-5.2729966e-08 -1.3901859e-12 -2.480062e-08 0.48020828402366866
-9.490284e-08 -1.5727615e-12 -4.4060755e-08 0.26806153846153846
-2.604904e-07 -1.38159e-11 -6.6738764e-08 0.3851277316097261

From Eddie:

Here's the recipe I used for apodizing and regularizing these for the unWISE catalog analysis:
https://github.com/schlafly/crowdsource/blob/6d9ae0681515df5c76c3afcd9843fd6a45a13bb3/python/wise_proc.py#L125
I don't remember how Aaron got the stamps exactly, but I remember thinking that the wings were noisy 
and that an overall small constant offset may not have been well constrained.