An examination of Figure 5 shows that the FSRQs in the DXRBS sample cover a
much wider range of parameter space than those in the two previously existing
complete samples of FSRQs, the 1 Jy and S4 samples
(radio data for these samples were taken from Stickel et al. 1994 and
Stickel & Kühr 1994, while X-ray data for these objects were taken from
the multifrequency AGN database of Padovani et al. (1997b; and references
therein);
it should be noted that
the deeper S5 sample is unidentified and half of the objects lack
redshifts, as noted in Stickel & Kühr 1996). We have quantified these
differences by 1-dimensional and 2-dimensional Kolmogorov-Smirnov (K-S) tests.
The 2-dimensional K-S test reveals that the differences in the (*L*_{X},*L*_{R})
plane coverage between the DXRBS and 1 Jy and S4 samples are significant: the
probability that the 1 Jy and DXRBS samples could emerge from the same parent
population is 0.2%, where as the probability that the S4 and DXRBS samples
could emerge from the same parent population is 2.2%. Given the fainter flux
limits of DXRBS, this is expected.

One-dimensional K-S tests reveal that largest difference is in the radio
luminosity. The probability that the 1 Jy and DXRBS radio luminosity
distributions could emerge from the same parent population is , and
the probability that the S4 and DXRBS radio luminosity distributions could
emerge from the same parent population is 0.7%. The mean of the DXRBS
*L*_{R} distribution is different from that of both the S4 and 1 Jy at greater
than 99.9% significance. The situation is somewhat different for the X-ray
luminosity distribution. The probability that the 1 keV luminosity
distribution of the 1 Jy and DXRBS sample could emerge from the same parent
population is 15%, i.e. our results are inconsistent with them emerging
from a different parent population. The result is similar for the S4 (23%
probability). Also, the mean of the DXRBS sample's X-ray luminosity
differs from that of both the 1 Jy and S4 samples' only at the
93 and level respectively. Note, however, that X-ray data are
available only for and of the S4 and 1 Jy FSRQs
respectively, so their X-ray luminosity distributions are likely to be skewed
towards the most luminous X-ray sources.

Inspection of Figure 5 reveals that the differences lie in two areas: at low luminosities (particularly low radio luminosities) and high ratios of . The former regime could not be surveyed well by previous surveys due to their considerably higher flux limits. It is therefore not surprising that, as shown in Figure 5, the 1 Jy and S4 samples together have only twelve objects at radio luminosities (), and none at . The fraction of low-luminosity objects is much higher in the DXRBS sample (Fig. 5), which, while still incomplete, already contains over twice as many objects (28; or ) with , six of which are at . The DXRBS sample is therefore the very first sample of blazars to contain statistically significant numbers of blazars at low luminosities, approaching what should be the lower end of the FSRQ luminosity function according to unified schemes, i.e. (Urry & Padovani 1995).

The discovery of a large population of FSRQs with ratios of X-ray to radio
luminosity *L*_{X}/*L*_{R} > 10^{-6} (), values more
similar to HBLs, is more startling, as few such objects were known in
previous complete samples (there are nine such objects in the 1 Jy and S4
combined; see Fig. 5). The finding of a large population
of ``HBL-type'' FSRQs contradicts the prediction of Sambruna et al. (1996)
that, based upon the similarities in the optical-X-ray broadband
spectral characteristics of LBLs and FSRQs, there should be no HBL-type
FSRQs. Padovani, Giommi & Fiore (1997b) were the first to
notice that about 17% of all radio quasars
with radio/optical/X-ray data (previous to DXRBS) fell in the region of the
plane typical of HBLs (or X-ray selected
BL Lacs) and called them ``HBL-like'' quasars.

We term these objects ``HFSRQs'', or
high-energy peaked FSRQs; this terminology stresses their apparent similarity
to the HBLs. These objects comprise of the DXRBS sample
of FSRQs so far. However they probably comprise a somewhat larger proportion
of the DXRBS FSRQ population as a whole, as (25/59) of the
newly-identified FSRQs are HFSRQs.
Padovani et al. (1997b) have proposed that the X-ray band in these objects,
unlike in lower *L*_{X}/*L*_{R} FSRQs, in which inverse Compton emission
prevails (Padovani, Giommi & Fiore 1997a),
is dominated by synchrotron emission (see also Sambruna 1997), as the
X-ray spectra of the previously-observed HFSRQs in their database were
as steep as those of HBLs (Perlman et al. 1996b; Sambruna et al. 1996;
Padovani & Giommi 1996). As the DXRBS
sample contains a larger, more representative sample of HFSRQs than could be
gleaned from previously identified samples, we will revisit this assertion
and address the properties of the HFSRQ subclass in depth in a future paper
(Perlman & Padovani, in prep). However, the data herein
allow us the first measure of the prevalence of such objects and their
proportion among FSRQs in a well defined sample, as well as the first
opportunity to speculate upon their relationship to the FSRQ subclass as a
whole.

In order to examine the differences between the HFSRQs and lower
*L*_{X}/*L*_{R} objects, we have performed 1-dimensional K-S tests on the radio
and X-ray luminosity distributions on the subsamples of DXRBS FSRQs with
*L*_{X}/*L*_{R} greater and less than 10^{-6}. These tests reveal that the
probability that the X-ray luminosity distribution of the two subsamples
could emerge from the same parent population is 41% (i.e. consistent with
having been drawn from the same parent population), while the probability
that the radio luminosity distribution of the two subsamples
could emerge from the same parent population is . The same story
is
told by the mean X-ray and radio luminosities: The mean radio luminosities
differ by 0.75 in the log and the significance of the difference is ,where as the difference in the X-ray luminosities is only 0.18, and is not
statistically significant (). This trend can also
be seen on Figure 5. There is only one HFSRQ at
luminosities , compared to over
two dozen lower-*L*_{X}/*L*_{R} objects. And a careful examination of the figure
reveals that the lower-*L*_{X}/*L*_{R} objects are much more strongly clustered
at high radio luminosities than are the HFSRQs. The fraction of HFSRQs also
increases as radio luminosity decreases. These trends are
similar to (but not as marked) as what is seen for BL Lacs (Urry & Padovani
1995; see also below).