Reliability

For this reliability test we used the coadds generated by the v3.2 pipeline for the 023* scans. The reliability was computed for 3 different coadds--information about the coadds can be found in the table below. The coadds were chosen to cover a large range of galactic latitudes, making certain there was a coadd in the galactic plane.

Coadds used for Reliability

Coadd ID	R.A. (J2000)	Dec. (J2000)	glon	glat
0208p062_023s_1_1	20.808306	6.264028	137.062920837161	-55.7329275014801
1602m590_023s_1_2	160.263663	-58.943815	286.796394306434	-0.218468482726562
2893p682_023s_1_1	289.374164	68.243125	99.4019532069342	22.7104698363461

The reliability is determined by taking the ratio of the coadd sources matched to truth sources to the total number of coadd sources (N_{MDEX->Truth matches}/N_{total MDEX sources}). A search radius of 3 arcsec is used. The MDEX source list is cut according to the following parameters:

• profile magnitude error <9.98 (w[1-4]sigmpro < 9.98)
• standard aperture error <9.98 (w[1-4]sigm < 9.98)
• number of deblended components = 1 (nb==1)
• source not actively deblended (na==0)
• SNR > 5 (W1,2,3); SNR > 2 (W4) (w[1-3]snr>5; w4snr>2)
• no saturated pixels (w[1-4]Sat == 0.0)
• nominmal photometry flags (w[1-4]flg == 0)
• coadd coverage above 4 to help with artifact removal (w[1-4]Cov)
• low profile fitting Chi square value (rchi2 <= 4)

The cuts are to ensure that the MDEX source list is as "clean" as possible. Early Reliability tests with this dataset showed that there many unflagged latents included in the source list. In lieu of a working latent artifact indentification module, I required the number of images contributing to a given MDEX object to be 4. This essentially eliminates the low-coverage crust of the coadd.

The following sets of plots show the cumulative reliability summed up from bright to faint sources. The vertical dotted line is the SNR=20 magnitude based on the science requirements. The horizontal dotted line denotes the 99.9% reliability limit, as stated in the science requirements. In addition to plotting the coadds individually, I've also plotted the total reliability by summing up the number of matches and MDEX sources in each magnitude bin (0.1 mags) over all three coadds.

The reliability at faint magnitudes varies quite a bit with sky position. In general, it becomes worse with increasing distance from the Galactic plane. For the most part, the Reliability requirement (which is for the Final Catalog) is not met for individual coadds at the shorter wavelengths.