2.7. Intrinsic Dispersion
As we will point out later, different quasar selection criteria produce different SEDs. Even with an homogeneous selection, the dispersion in the SEDs of AGN is rather large, about an order of magnitude in the IR and UV, even when normalized at the 1 µm "inflaction point". The dispersion in the SED of local X-ray selected quasars is emphasized in E94 and shown in Figure 8, but it is often not taken into account.
Figure 8. Intrinsic dispersion in the SED of E94. Curves represent the 99%, 90%, and 68% dispersion with respect to the best fit SED and are normalized in order to have the same flux at 1.25 µm.
We believe this dispersion to be a fundamental property of quasars that always should be considered when referring to an average SED. The reason is both "physical", in order to have a physically correct view of quasar emission, and "computational", since the results obtained when computing integral properties of quasar samples can be significantly altered by a non-zero dispersion distribution of parameters. An example is the correction from the observed to the "effective" OX discussed in Section 2.5.
We already discussed the dispersion in the IR emission in Section 2.7. Here we only note that, even if part of the observed dispersion (see Fig. 6) is due to a star formation contribution, it is likely that the intrinsic dispersion in the IR-to-bolometric ratio is a factor of ~ 2. Finally, in Figure 9 we show the distribution of OX for the sample of optically selected quasars observed with ROSAT (Yuan et al. 1998). Approximating the distribution with a Gaussian, the standard deviation is (OX) ~ 0.2, corresponding to a dispersion in the ratio between optical and X-ray emission of a factor of ~ 3.
Figure 9. Distribution of OX for a sample of ~ 1000 optically selected quasars observed with ROSAT. Empty histogram shows the ROSAT detections; shaded histogram takes into account upper limits, as well. Figure from Yuan et al. (1998; their Fig. 9).