---

High-Chisqr in One out of Three Bands

---

The following query was run for each band to detect high quality three-band detections with single-band outliers in chisqr. If the hot pixel problem contaminates stars then there could be a population of sources in the database with corrupted photometry. This analysis focuses only on AAA sources.

create table hih as select * from v3_pscat where ph_qual='AAA' and
h_psfchi between 2.5 and 10.0 and j_psfchi<1.3 and k_psfchi<1.3 and
prox>8. and cc_flg='000'

   and similarly for J and Ks...

Counts -- total and abs(glat)>20.:

wsdb=> select count(*) from hij; ----  57618
wsdb=> select count(*) from hih; ----  38298
wsdb=> select count(*) from hik; ---- 100540

wsdb=> select count(*) from hij where abs(equ2glat(ra,decl))>20.;  11323
wsdb=> select count(*) from hih where abs(equ2glat(ra,decl))>20.;  10259
wsdb=> select count(*) from hik where abs(equ2glat(ra,decl))>20.;  21641

Below are color magnitude diagrams for the three populations. Observations:

• The H-band does show a haze of oddly colored sources suggesting contamination from the hot-pixel problem. This haze is about 1% of the total population, so roughly a couple of hundred sources out of the entire database. J-band may also show a slight (non-physical) haze to the right of the main-sequence locus.
• Ks-band produces a more clean diagram. Oddly there seems to be a bias toward bright sources.
• The excessive number of Ks-band sources is consistent with these sources being "clean" in J- and H-band, but having contaminated profiles when more stars show through the dust at Ks-band, however this does not explain the similarly excessive number at high latitude.

AAA Sources with large chisqr in one band at High-latitude

J-band	H-band	Ks-band