search

msimet / Stile

Stile: the Systematics Tests In Lensing pipeline
BSD 3-Clause "New" or "Revised" License
9 stars 6 forks source link

Updating HSC flags #66

Open msimet opened 9 years ago

msimet commented 9 years ago

We just had a discussion about which flags we should actually be using in the HSC module--this is a reminder to make these updates.

msimet commented 9 years ago

Okay. Here's a list of the flags we discussed, and the decisions we made:

to keep as standard Stile flags:deblend.nchild (on single exposures only--it's included in some of the other coadd flags already), flags.badcentroid, deblend.skipped, centroid.sdss.flags, flags.pixel.edge, shape.hsm.regauss.flags,cmodel.flux.flags, detect.is-primary (on coadds), flags.pixel.cr.any (on single exposures)

to remove as standard Stile flags: deblend.too-many-peaks, deblend.parent-too-big, deblend.has.stray.flux, flags.pixel.interpolated.any, shape.sdss.centroid.flags (doesn't propagate into measurements--it might tell you something is wrong, but something else probably would do that first), flags.negative, centroid.naive.flags, flags.pixel.bad

To get rid of all deblended objects (which we might want to do for testing purposes), those cuts would be parent = 0 and deblend.nchild = 0.

I also have a note about discussion of PSF measurement flags, but I didn't write down any action items for that discussion.

Finally, there were a few flags whose discussion I either didn't write down, or didn't understand the conclusions we came to: deblend.failed, flags.pixel.interpolated.center, flags.pixel.saturated.any, flags.pixel.clipped.any, flags.pixel.saturated.center, flags.pixel.cr.center, flags.pixel.suspect.center. I think the conclusion was that the .any flags were not necessary but the .center ones we should keep, but I wasn't sure; for deblend.failed I'm afraid I don't remember at all.

Folks in the room: does this seem a reasonable summary of what we discussed? Did I miss something? Is the .any vs .center distinction the one we actually decided on?

TallJimbo commented 9 years ago

That matches my recollection. I do think you got the .any vs. .center distinction right too.

I'm afraid I don't actually know what deblend.failed does, but it does sound scary. Do we know if it's actually being set on any objects.

msimet commented 9 years ago

I haven't checked if deblend.failed is set on anything--will investigate.

rmandelb commented 9 years ago

@msimet , could you also make sure this is linked on the HSC internal wiki page where we summarized the discussions from our meeting?

HironaoMiyatake commented 9 years ago

We should keep flags.pixel.clipped.any. This "any" is different from other flags; this "any" means there is some exposures clipped by sigma-clipping in the coadd process, and the other "any" means that there is some bad pixel (saturated, cr, etc...) in outskirts of objects (outside of 3x3 pixels around the center, if I remember correctly).

HironaoMiyatake commented 9 years ago

Why do not we need flags.pixel.bad?