Match Sprinkled AGN on magnitude

jbkalmbach commented 8 years ago

Currently only matching source AGN on redshift. Need to add a requirement on magnitude as well.

drphilmarshall commented 8 years ago

i-band mean AB magnitude to be exact (which OM10 is not).

jbkalmbach commented 8 years ago

Hi @drphilmarshall, the columns in OM10 I'm wondering about are:

MAGI_IN
MAGI
MAG (From previous discussion I know this is magnification, but just want to verify)
ABMAG_I (I figure this is absolute and the one below is apparent, but magnitude of what?)
APMAG_I

drphilmarshall commented 8 years ago

Apologies for the nightmarish naming scheme - they were not my choices!

MAGI_IN = source plane (ie unlensed, intrinsic) i-band AGN mean magnitude
MAGI = i-band magnitude of 3rd brightest (in a quad) or 2nd brightest (in a double) AGN image
MAG - magnification factors for each AGN image (ie MAGI = MAGI_IN - 2.5*log10(MAG[k]) where k is in the index of the 3rd (or 2nd) brightest image)
ABMAG_I = Estimated absolute magnitude of the lens galaxy, in the (rest-frame) i-band
APMAG_I = Estimated i-band apparent magnitude of the lens galaxy
VELDISP = velocity dispersion of the lens galaxy, in km/s

I do not trust the ABMAG_I and APMAG_I: better to find a massive CatSim galaxy that has (approximately) the same redshift and velocity dispersion, and use that.

jbkalmbach commented 8 years ago

Ok, so what I have put in place now after learning this is: 1) MAGI_IN values appear to be apparent magnitudes not the AB Mags you mentioned matching to above so I will match using apparent agn magnitudes as well. 2) Convert all om10 MAGI_IN values from i-band to imsimband to get mag_norm values. 3) Match sprinkled objects according to redshift +/- 0.05 and now additionally match agn_magnorm values to converted MAGI_IN values +/- 0.25 (what should this be?)

And then the final step will be adjusting the density parameter to account for the new matching requirements once we decide on a matching range for magnitude.

What do you think @danielsf @rbiswas4 @drphilmarshall?

jbkalmbach commented 8 years ago

Code visible in branch: https://github.com/DarkEnergyScienceCollaboration/Twinkles/tree/issue/309/match_sprinkled_agn_on_magnitude

rbiswas4 commented 8 years ago

@jbkalmbach

Is the reason that you prefer matching in mag rather than velocity dispersion because you don't have velocity dispersion on the catsim galaxies? From a different thread, I think @drphilmarshall explained that velocity dispersion should be a proxy for stellar mass, but I suppose you cannot do it without the exact mapping. Would it be easy to do it fi someone provided you a function

def mass_stellar(vel_disp):
     ....

or would that still be harder than matching the i band magnitudes?

jbkalmbach commented 8 years ago

This issue is for the AGNs. I think we are planning on using some sort of mapping between vel. disp. and mass for the lens galaxies.

rbiswas4 commented 8 years ago

Ah ok! So that is still the plan.

drphilmarshall commented 8 years ago

Your plan as written sounds good, Bryce. Just as point of information, "AB" magnitudes are apparent mags in the AB system (ie relative to a flat spectrum of standard normalization in Jy). An absolute magnitude can also be given in the AB system and are the apparent magnitudes that you would measure if you were magically lucky/unlucky enough to be just 10pc away. I was confused between AB and absolute for months when I switched to doing optical astronomy. It is also by no means the only confusing thing, I found :-)

PS. I think 0.25 mags slop in matching the AGN mags sounds sensible.

jbkalmbach commented 8 years ago

The 0.25 mags slop parameter gives us 132 lensed AGN systems when the density parameter is set at 1.0 in the sprinkler. Do we want to increase this leniency in magnitude to get more systems or are we happy with that number?

drphilmarshall commented 8 years ago

It'd be good to get to 400, to match the number expected from TDC1 (and to give us a good chance of getting ~100 well measured time delays). Let us know what it takes!

On Thu, Sep 22, 2016 at 11:47 AM, Bryce Kalmbach notifications@github.com wrote:

The 0.25 mags slop parameter gives us 132 lensed AGN systems when the density parameter is set at 1.0 in the sprinkler. Do we want to increase this leniency in magnitude to get more systems or are we happy with that number?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DarkEnergyScienceCollaboration/Twinkles/issues/309#issuecomment-248993100, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY93EisidEUoOez4NePIzPnQwTRCSPks5qss1BgaJpZM4J4W8t .

jbkalmbach commented 8 years ago

Expanding the magnitude bound to +/- 1.5 and the redshift to +/- .1 gives 352, how's that?

drphilmarshall commented 8 years ago

Hmm - that's pretty loose in magnitude, they need to be bright enough to see! Can you try widening in redshift at fixed 0.25 mag spread please? Sorry, should have thought of this earlier.

On Fri, Sep 23, 2016 at 3:23 PM, Bryce Kalmbach notifications@github.com wrote:

Expanding the magnitude bound to +/- 1.5 and the redshift to +/- .1 gives 352, how's that?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DarkEnergyScienceCollaboration/Twinkles/issues/309#issuecomment-249316528, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY93yAeYktBmts6F-zbvxxk_lC_357ks5qtFF5gaJpZM4J4W8t .

jbkalmbach commented 8 years ago

Ok, with a fixed .25 mag spread I did a few runs with different redshift spread.

Redshift Spread	# Systems
.5	264
.75	307
1.0	353

Let me know where you think the best tradeoff occurs and we'll set that up.

drphilmarshall commented 8 years ago

Even 0.5 is making me nervous. Can you try searching in a percentage range rather than an absolute one, please? I worry that zs = [2.5:3:5] is probably OK (as in the corresponding change in image separation and time delay is not large) but zs = [0.5 : 1.5] is not, because the image separations will come out wrong. I'd like to see what 10%, 25%, 50% ranges get us. Actually, since the 0.25 mag range is 0.1 dex, or about 30% in flux, you could try searching in uniform ranges (0.1, 0.2, 0.25, 0.3) in log10(z) in the same way. Thanks!

BTW this procedure will lead to simulated systems where the source redshift does not match the observed image separation and time delay: it'll be as if the lens galaxies have unexpected scatter in their masses, and they may end up looking systematically different from a population of singular isothermal ellipsoids. While this is not a problem for Twinkles 1, where we are only interested in the deblending of images and extraction of light curve and measurement of time delay, it would be a problem if we tried to do those things using a hierarchical model for the populations of lenses and sources - so we must aim to do better in Twinkles 2.

On Mon, Sep 26, 2016 at 9:59 PM, Bryce Kalmbach notifications@github.com wrote:

Ok, with a fixed .25 mag spread I did a few runs with different redshift spread. Redshift Spread # Systems .5 264 .75 307 1.0 353

Let me know where you think the best tradeoff occurs and we'll set that up.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/DarkEnergyScienceCollaboration/Twinkles/issues/309#issuecomment-249766703, or mute the thread https://github.com/notifications/unsubscribe-auth/AArY98XjJtotp3pEVUb-4kv7lXzgT0sKks5quKLFgaJpZM4J4W8t .

jbkalmbach commented 8 years ago

Results with magnitude difference fixed at .25:

Redshift Difference (dex)	# Systems
.1	199
.2	260
.25	281
.3	306

drphilmarshall commented 8 years ago

Let's go with 0.1 dex in both zs and flux then (=0.25 in magnitude). A sample of 200 lenses should make for a good analysis. Thanks Bryce! :-)

jbkalmbach commented 8 years ago

With merging of #336 we can close this.

LSSTDESC / Twinkles

Match Sprinkled AGN on magnitude #309