The results now allow us to evaluate the original selection criteria and formulate new improved criteria for future observations. The two primary selection criteria were the radio/optical position agreement and the APM stellar/nonstellar classification. Of the 219 candidates, 97 were classified as stellar on both plates, 29 as stellar on the E plate only, and 93 as stellar on the blue plate only. Of the 69 confirmed QSOs, the three comparable numbers are 55, 3 and 11. Clearly these numbers leave open the possibility that some QSOs are classified as nonstellar on both plates; we will return to this issue below. The reliability of the APM classifier is magnitude dependent: all 11 QSOs which are stellar on the O plate only are 18th magnitude or brighter on the O plate. Seventy-four fainter candidates are included in the initial sample of 219 because of a stellar classification from only the O plate; all but 6 of these have O-E colors redder than 2.0. Thirty of these have spectroscopic classifications and are all either ELG or ALG. In the continuation of the survey these can be eliminated by selecting objects with O-E < 2.0 (see below). A color magnitude diagram of all the survey objects with redshifts and spectral classifications is shown in Figure 6.
Figure: Color magnitude diagram of the entire spectroscopic sample. The reddest QSO has O-E = 1.74; emission line galaxies have colors intermediate between the QSOs and absorption line systems.
The angular separation selection criterion turned out to be too generous. Of the 69 QSOs in the sample, 64 have separations between the radio position and optical counterpart of < 1."0, 3 between 1."0 - 1."1, and 3 between 1."1-2."0. The 2 QSOs with the greatest separation are extended radio sources. Of the original 219 candidates, over 25% lie outside 1."1, so a tighter matching radius will eliminate many false candidates at the expense of only of the QSOs.
Although color was not used in the original selection, the survey is not finding very red quasars: only five QSOs in the sample are redder than 1.5, the reddest with O-E = 1.74. This is perhaps because of the relatively bright magnitude limit; typical highly reddened QSOs will be much fainter in the optical (Webster et al. 1995) and will not appear in our candidate list. Imposing a color cut on our candidate sample will eliminate a sizable fraction of the non-QSO objects while sacrificing at most only a tiny number of QSOs, albeit potentially interesting ones. In the whole sample, 104 candidates out of 219 are redder than 1.75. We classified 63 of these objects spectroscopically; 34 as galaxies, 18 as ELG, 11 as Galactic stars, and none as QSOs. For an additional test of using a color cut in selecting candidates, we matched the Véron-Cetty & Véron (1996) QSO catalog against the APM catalog. Of the 380 QSOs that were within 1" of an APM object brighter than 17.5 on the E plate, only one was redder than 2.00, confirming the utility of color as a selection criterion.
A more efficient observing program, then, would restrict candidates to those optical counterparts classified as stellar on either POSS plate, falling within a 1."1 matching radius, and with colors bluer than E-O = 2.0. If the pilot survey is typical, such a sample would be 70% QSOs and 95% complete. Only two of the 68 candidates still without spectroscopic classification in the current sample would survive the new selection criteria.
Two potential causes of incompleteness are the absence of a core radio source in a radio loud QSO, and a nonstellar classification on both POSS plates. To estimate the magnitude of these two effects, we have carried out several tests. As mentioned above, the Véron catalog of QSOs was matched with the APM catalog, selecting objects with E < 17.5 and separations ; 380 matches resulted. Of these 380 QSOs, 27 (7%) were classified as nonstellar on both POSS plates; however, only 6 ( ) had M brighter than -24.3, implying that most of these are Seyfert 1 galaxies and are probably correctly classified as nonstellar.
It is more difficult to estimate the fraction of radio loud QSOs without core components. We have inspected the FIRST survey images around the positions of all the Véron QSOs in the survey area and have found one radio loud QSO (B2 1248+30) with E = 17.5 that our selection criteria missed because it has no core radio component. This suggests that few QSOs are missed because of this effect, but the test is inconclusive because other radio-selected QSO surveys may suffer from a similar incompleteness.