Report for Simulation #5

The bottom line is that we get 99% completeness, with all of the missing QSOs in the 2.5 - 3.5 redshift range. In this range, the completeness drops to 88%. The efficiency is 53%. Of the selected candidates, 53% are QSOs, 14% are NELGs, 4% are white dwarfs, and 30% are galactic stars. This assumes that the simulations are a fair representation of the sky, and that I didn't cheat too much by using the same data to fit the locus.

The new algorithm handles stars which are not detected in all bands. However, it needs a training set of stars which both covers the entire color range and has all colors for the stars. The red stars in Xiaohui's simulation do not have u measurements, and presented problems for creating the locus fit.

I have (more or less) successfully run the fifth simulation through the QSO target selection code. I was able to twiddle the parameters to get about the same efficiency and completeness that I got before we put the red stars in, but I have a few reservations about the fitting procedure. This simulation is different from previous simulations in that Xiaohui put in a limiting magnitude (mostly in the u' filter) instead of the simulated magnitude if the magnitude was fainter than our detection limit on the 2.5 meter. The selection algorithm has been modified to recognize an object as a star if there is any way to extrapolate it into the locus, given the known limits in its colors. For instance, if the object is not detected in u, then we know that u-g > ulimit-g. If we can find any value of u-g in this range which falls inside our stellar locus fit, then it is assumed to be a star.

I had a few problems running this simulation through; the problems point out the need for special photometric data to tune the locus parameters.

I could not separate the stars out by magnitude, and make a different locus for each magnitude range. This is because for the fainter stars the redder stars were limits. So, I could not fit the red end of the locus. I need the red end of the locus to be in the correct position so that when I project the limits back to the locus the projected point is within the locus width. So, I generated only one locus fit, using stars at 15 Because of these problems, I editted the locus parameters by hand to make the red end very wide (and thus catch most of the badly extrapolated red points in the web). I also editted the blue end to make it slightly narrower, to find more z ~ 2.5 QSOs (well, I found one more, at any rate). I'm not quite sure this last edit was justified, but the end of the locus was wide since I could not separate by magnitude.

This is the figure of merit:

                                        # targets/#objects brighter than 20. # objects

Targets selected (by type):     1       41/6558 6558
                                2       78/4700 4700
                                3       0/0     0
                                4       0/0     0
                                5       0/0     0
                                6       0/0     0
                                7       0/0     0
                                8       57/58   58
                                9       14/16   16
                                10      211/213 213
                                        -----------------
                                        401/11545       11545

QSO completeness (z):           -10 - 1 85/85   85
                                1 - 1.5 49/49   49
                                1.5 - 2 32/32   32
                                2 - 2.5 24/24   24
                                2.5 - 3 14/16   16
                                3 - 3.5 6/6     6
                                3.5 - 4 1/1     1
                                4 - 6   0/0     0
                                6 - 10  0/0     0
                                        -----------------
                                        211/213 213

QSO completeness (m):           -10 - 15        0/0     0
                                15 - 16 1/1     1
                                16 - 17 1/1     1
                                17 - 18 20/20   20
                                18 - 19 91/92   92
                                19 - 20 98/99   99
                                20 - 30 0/0     0
                                        -----------------
                                        211/213 213