LOW_QUALITY SV target class

[changhoonhahn]

We are currently contemplating two different scenarios for testing the nominal quality cuts:

1. LOW_QUALITY target class only includes objects that fail the color cuts. Objects that fail the FRAC_FLUX_*, FRACMASKED_*, and FRACIN_* are included in the other target classes.
2. LOW_QUALITY target class includes objects that fail the color cuts and objects that fail the FRAC_FLUX_*, FRACMASKED_*, and FRACIN_*.

We should pick the scenario that give us enough objects to test:
Scenario 1:
N_colorfail = (60/deg^2 requested) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 6750
N_fracfail = (34/deg^2) x (540/850 BRIGHT subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) + (19/deg^2) x (300/750 FAINT subsampling) x (0.5 fib. assign fraction) x (30 fields) = 3285
Scenario 2:
N_colorfail = (30/deg^2 requested) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3375
N_fracfail = (30/deg^2 requested) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3375

questions:

• @qmxp55 Are the numbers above correct?
• Are more concerned about objects that fail the color cut or objects that fail the FRAC* cuts? Scenario 1 has more color failed objects; scenario 2 has more FRAC* failed objects.

[qmxp55]

@changhoonhahn, where do you get the 15.53 and 8 in N_fracfail of Scenario 1? Shouldn't those be 34 and 19? as in the red box of the third slide: https://docs.google.com/presentation/d/1UYTyjki9vz_FCk0Ks_xjSRwUXZq4OhVu1IAotc6oXrI/edit?usp=sharing

In the second scenario, I think a more realistic view will be:
N_colorfail = (47/deg^2) x (60/100 LOW Q subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3172
N_fracfail = (53/deg^2) x (60/100 LOW Q subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3577

[qmxp55]

@changhoonhahn, the N_colorfail also includes the NOBS fails.

[changhoonhahn]

@qmxp55 Yes they should be. 15.53 and 8 were from faulty memory.

So with Scenario 2, we get ~300 more (+10%) FRAC* failed objects but ~3500 fewer (-50%) color cut+nobs failed objects.

[qmxp55]

@changhoonhahn I'm not sure about N_colorfail = (60/deg^2 requested) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 6750, That's the goal, however, if only NOBS and COLOUR CUTS, it should be the same as in Scenario 2: N_colorfail = (47/deg^2) x (60/100 LOW Q subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3172.

[qmxp55]

Summarizing.

LOWQ sample in scenario (1) does include samples of CC and NOBS cuts, while in scenario (2) in CC, NOBS and FRACS* (i.e. FRAC_FLUX_*, FRACMASKED_*, and FRACIN_*). The FRACS* sample is around 37 objects per sq. degree, where FRACS* in (1) will come in the BRIGHT and FAINT selections as follows:

Scenario 1
FRACS* in BRIGHT = 21
FRACS* in FAINT = 16
N_fracfail = (21/deg^2) x (540/880 BRIGHT subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2)
+ (16/deg^2) x (300/790 FAINT subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 2130

and in (2) in the LOWQ:

Scenario 2
FRACS* in LOWQ = 37
FRACS* in FAINT = 16
N_fracfail = (37/deg^2) x (60/64 BRIGHT subsampling) x (0.5 fib. assign fraction) x (30 fields) x (7.5deg2) = 3900

Therefore, we have decided to go for Scenario 2 because we get almost twice the sample than in Scenario 1.

[qmxp55]

Latest updates pushed to DESITARGET in #659