\documentstyle[12pt,aasms]{article}
\pagestyle{plain}%  | {empty} | {headings}
\topmargin=-0.3in
\oddsidemargin=0.2in
\textwidth=6.0in
\textheight=8.5in
\parindent=3.0em	% Depends on type size
\headsep=0.45in
\parskip 6pt plus 1pt	% Extra space between paragraphs
\footnotesep \baselineskip	% Separate footnotes from text
\sloppy
\pretolerance=500 \hyphenpenalty=200   % Resist hyphenation
%
\newcommand{\etal}{{\em et~al.}}
\newcommand{\lapprox}{_<\atop^\sim}  % math mode only!
\newcommand{\gapprox}{_>\atop^\sim}  % math mode only!
\newcommand{\smb}{3000{\rm\AA}~bump~}
\newcommand{\bump}{bump~}
\newcommand{\xem}{\rm}
\newcommand{\munorm}{1.25}
\newcommand{\bumps}{bumps~}
\newcommand{\bbump}{Big Bump~}
\newcommand{\wbump}{weak bump~}
\newcommand{\wbumps}{weak bumps~}
\newcommand{\nqso}{47} % Size of UV sample
\newcommand{\nqrq}{29} % No of radio-quiets
\newcommand{\nqrl}{18} % No of radio-louds
\newcommand{\npg}{27}  % No of PGs
\newcommand{\nsy}{8}  % No of Sys
\newcommand{\nhost}{29} % No with host data
\newcommand{\nnohost}{18} % No with host data
\newcommand{\nother}{23} % No in table 1b
\newcommand{\vvs}{\vskip 0.2in\par\noindent}
\newcommand{\resolve}[1]{{#1}}
%
\newcommand{\tsz}{\footnotesize}
\newcommand{\er}{\mbox{$\pm$}}
\newcommand{\sam}{UVSX sample} % 
\newcommand{\usam}{UVSX SAMPLE} % 
\newcommand{\uother}{OTHER ``IPC''} % 
\newcommand{\other}{other ``IPC''} % 
%
%
\begin{document}
\doublespace
\title{ATLAS OF QUASAR ENERGY DISTRIBUTIONS\altaffilmark{1}\\
 I. THE DATA }
\author{
Martin Elvis\altaffilmark{2}, 
Belinda J. Wilkes\altaffilmark{2},
Jonathan C. McDowell\altaffilmark{2},
Richard F. Green\altaffilmark{3}, Jill Bechtold\altaffilmark{4},
S. P. Willner\altaffilmark{2}, M.S. Oey\altaffilmark{2,4},
Elisha Polomski\altaffilmark{2,5}, and Roc Cutri\altaffilmark{4}}

\altaffiltext{1}{Based in part on data acquired with the {\it International 
Ultraviolet Explorer}~satellite, operated at the Goddard Space Flight Center
for the National Aeronautics and Space Administration, the Multiple Mirror
Telescope (MMT), 
a joint facility of the Smithsonian Institution and the University
of Arizona, and the Infrared Telescope Facility (IRTF), a joint facility of NASA
and the University of Hawaii.}
\altaffiltext{2}{Center for Astrophysics, 60 Garden Street, Cambridge, MA 02138}
\altaffiltext{3}{National Optical Astronomy Observatories,
Kitt Peak National Observatory, Tucson, AZ 85726. KPNO is
operated by the Association of Universities for Research in
Astronomy, Inc., under contract with the NSF.}
\altaffiltext{4}{Steward Observatory, University of Arizona, Tucson, AZ 85721}
\altaffiltext{5}{Center for EUV Astrophysics, Berkeley, CA}

\date{Version: \today \\
\small To be submitted to {\it Astrophysical Journal Supplement Series}}



\begin{abstract}
We present an atlas of the spectral energy distributions (SEDs) of 
normal,  non-blazar, quasars over the whole available range (radio to 10 keV
X-rays) of the electromagnetic spectrum. 
The primary (UVSX) sample includes \nqso~quasars for which
the spectral energy distributions include X-ray spectral indices
and UV data. Of these
\nqrq~are radio-quiet and \nqrl~are radio-loud.
The SEDs are presented both in figures and in tabular form, with
additional tabular material published on CD-ROM.
Previously unpublished observational data for a second
set of quasars excluded from the primary sample are also tabulated.
The effects of host galaxy starlight contamination and 
foreground extinction on the \sam~are considered and the sample is used
to investigate the range of SED properties. Of course, the
properties we derive are influenced strongly by the selection
effects induced by quasar discovery techniques. We derive the mean
energy distribution (MED)
for radio-loud and radio-quiet objects and present the
bolometric corrections derived from it. We note however that
the dispersion about this mean is large ($\sim$1 decade for
both the infrared and ultraviolet components when the MED is
normalized at the near infrared inflection). At least part of
the dispersion in the ultraviolet may be due to time variability,
but this is unlikely to be important in the infrared.
The existence of such a large dispersion indicates that the MED reflects
only some of the properties of quasars, and so
should be used only with caution. 
\end{abstract}

\keywords{Galaxies: Quasars: General --- Atlases}


\section{Introduction}

\bigskip

One of the reasons that main sequence stars are much better understood
than quasars is that they radiate (almost) black body spectra with
temperatures between $\sim$4000 and $\sim$30,000~K, so that their 
black body peak moves
conveniently through the optical band. The resulting strong color changes
allowed the early recognition of the main sequence in the Hertzprung-Russell
diagram. By showing that most stars lie in a restricted band of
color and luminosity this diagram provided a crucial input to theories
of stellar structure. The current lack of understanding of quasars may
correspondingly be due to the distribution of their continuum light. 
Because the quasar phenomenon covers an extremely broad range of wavelengths
it is hard to see continuum features analogous to the black body peak in
normal stars. Quasars 
emit almost constant power per
decade of frequency from 100 $\mu$m to at least 100 keV (see \eg\ figure 1
of Carleton et al., 1987). While this equipartition is surprising, and may
be to some extent an observational artifact (Elvis and Brissenden
1993, in preparation), 
 it contains too
little information to constrain theoretical ideas.
Overcoming this problem requires the assembly of  Spectral Energy
Distributions (SEDs) for sizable samples of quasars
over the whole accessible range of the electromagnetic spectrum
from radio to hard X-rays (and in the future, gamma rays).

In this paper we present SEDs for a sample of \nqso~quasars.
We concentrate on `normal' quasars with little or no
polarization and no dramatic optical variability. Our primary
selection criteria were: (1) existing \ein\ observations at good
signal-to-noise (to ensure good X-ray spectra); and (2) an optical
magnitude bright enough to make an IUE spectrum obtainable.
Making the primary selection at these wavelengths means
that not all the sample are detected in the IRAS data.

Several previous studies of quasar SEDs have been published.
Each emphasized a particular region, and none considered the X-rays
in detail. Early work included SEDs for IRAS-bright AGNs by
Edelson and Malkan (1986), and for hard X-ray selected Seyferts
by Ward et al (1987).
Infrared to optical SEDs for the PG quasars were presented by Sanders et al.
(1989) with an emphasis on explaining the infrared. Near-infrared to
ultraviolet SEDs for IUE observed quasars were presented by Sun and Malkan
(1989). 
Our study is different in that it includes X-ray data,
divided into three energy bands in many cases,  for all the objects, 
and includes
both IUE and IRAS data. The X-rays are important to define
the total luminosity, and are crucial in overall modelling as 
they seem to come most directly from the 
central source.
The sample is fairly evenly divided between radio-quiet and radio-loud
objects.

We present the data in \S 4, and the necessary corrections for reddening,
variability and, importantly, host galaxy contributions in \S 5.
Section 6 discusses the properties of the SEDs, including their mean
and dispersion. First, however, we outline the features we
are studying (\S 2) and the sample of quasars (\S 3).

\section{Continuum Features}

Fig. 1  shows the radio to X-ray rest frame energy distributions 
of two typical quasars, 4C 34.47 and Mkn586. All energy distributions
are plotted as log $\nu$f$_\nu$ vs. 
log $\nu$;
since $\nu$f$_\nu$ is the flux per logarithmic
frequency interval,
such plots give the best indication of the frequency ranges
where most energy is released. The X-ray data are indicated with a `bow-tie' symbol
representing a power law fit with both the best fit slope and the 1 sigma
confidence limit slopes indicated. In general, no data are available in 
the extreme 
ultraviolet `gap' beyond the Lyman limit where our galaxy is opaque.

In the 1-100 $\mu$m 
infrared band both quasars are almost flat (Ward et al. 1987, Neugebauer et al. 
1987, hereafter N87). A single, nearly horizontal, power-law fits the IR points 
reasonably well and intersects 
the X-ray point at about 1 keV.
We will call deviations from this power-law `continuum
features'. There are four prominent features of this kind:

\begin{enumerate}
\item The power output always drops in the sub-millimeter band
( the ``mm-break'', Fig. 1) but the size of the drop varies dramatically
from object to object. Quasars in which the drop is only 2 decades are
called ``radio-loud'' (e.g. 4C 34.47). The great majority of quasars have a
much stronger mm-break of 5 or even 6 decades 
(Condon et al 1981, Kellermann et al. 1989)
and are called ``radio-quiet''
(e.g. Mkn586). This distinction between radio-loud 
and radio-quiet quasars
is the oldest in the quasar literature and goes back to the `blue
interlopers' found in early radio source identification 
work (Sandage 1965).
Radio-quiet objects are much the more
common, by about a factor 10 (Kellerman et al 1989). 
The mm-break is the strongest known feature in normal quasar
continua, although it is less strong in objects selected
at high radio frequencies. 

\item The optical-ultraviolet continuum rises above the infrared
and forms a ``UV bump'' (Shields 1978, Malkan and Sargent 1982,
N87). 
Variability studies show that this is a separate
component from the infrared (Cutri et al. 1985) since it varies
much more strongly.
This big bump is most often interpreted in terms of thermal
emission from an accretion disk ( e.g. Malkan 1983, Czerny and Elvis 
1987). The strength of the feature in our study may be affected by selection
effects since many quasars were discovered by ultraviolet excess techniques.
The beginning of the bump is marked by an inflection between
1 and 1.5 $\mu$m in the rest frame; this ``near infrared inflection'' is the only continuum
feature whose wavelength is well defined (N87). 

\item X-ray spectra of many radio-loud quasars and high luminosity
Seyfert 1 galaxies have rising slopes toward high frequencies in log $\nu$f$_\nu$ 
vs. log $\nu$
energy distributions (Zamorani et al 1981,
Mushotzky 1984, Bezler et al. 1984, Turner and Pounds 1989,
Williams et al 1992). 
They cannot then be an extension of 
a flat, or slightly falling, infrared power-law, as
has been suggested for some objects by Carleton et al (1987). A new emission 
component must be emerging in the X-rays above 1 keV 
in these objects. 

\item The most recently identified continuum feature is the `XUV excess' (e.g. 
Arnaud et al 1985, Wilkes and Elvis 1987 (WE87), Turner and Pounds 1989, Masnou
et al 1992).  The 1 keV spectrum of quasars is often fitted with a power
law spectrum since the available spectral resolution is very low;
studies using the {\it Einstein}~IPC and EXOSAT showed that excess flux
above this power law was often present in the ultra-soft (0.3 keV and
below) region of the spectrum; the excess can be highly
variable (e.g. Turner and Pounds 1988,
Elvis et al 1991).  It is possible that the XUV excess is
the same physical component as the ultraviolet bump since the energy
distribution rises towards the EUV from both sides of the unobserved
spectral region. WE87 found evidence
for a soft excess in 8 of their 33 quasars. Subsequent reanalysis
of the 13 highest S/N quasars in this sample by
Masnou et al (1992) detected a soft excess in about half
of the objects. A similar fraction was 
found by Turner and Pounds (1989) for Seyferts.

\end{enumerate}


\vspace{0.1in}

\section{The \sam}

We have carried out an extensive program of quasar continuum
observations over the past seven years, the results of which
we present here. Since we believe it important to include X-ray observations,
we restricted ourselves to objects that had been observed 
to high S/N with the Imaging Proportional Counter (IPC, Gorenstein, Harnden and
Fabricant 1979)
instrument on the {\it Einstein Observatory}~(HEAO 2, Giacconi et al 1979) satellite.
We note that this introduces a bias towards objects which are low redshift,
moderate luminosity
and have strong X-ray to optical flux ratios (WE87); the sample is heterogenous
and not flux-limited.
The objects were selected mainly from the PG (Schmidt and Green 1983),
 3C and Parkes catalogs. No
single completeness criterion could be used, however, because of the
quasi-random way in which the original \ein\ observations were made.

We have selected a subsample of \nqso~quasars which have sufficient counts (${\gapprox~}~300$) in the 
IPC  to give a reasonably constrained power law spectral fit, and which are
optically bright enough (V$<17$) to be observable with IUE: we designate 
this the `\sam'. 
\nqrq~members of the sample are radio-quiet and \nqrl~are radio-loud.
The X-ray and ultraviolet data have been combined with IRAS space-based
infrared data and ground-based optical, infrared and radio observations to construct
the complete energy distributions. 30 of the \nqso~objects are detected
by IRAS at 60 microns; only 15 are detected at 100 microns.

The \nqso~quasars in the \sam~are listed in Table 1(a). 
The sample has substantial overlap with 
the work of Elvis et al (1986) and WE87. The table gives
the common name of the quasar as well as its catalog number in
the {\it Einstein Observatory Source Catalog}~(Harris et al 1991),
and the name of the associated host galaxy where appropriate.
Both B1950 and J2000 coordinates are provided, as well as the 
redshift and typical V magnitude. The estimate of the foreground
Galactic hydrogen column density (in units of $10^{20}\mbox{ cm$^{-2}$}$)
is listed together with a reference. Most of the column estimates
are accurate values from Elvis, Lockman and Wilkes (1989).
Finally, each object is given a classification. The classifications are
as follows: radio-quiet (RQ), broad absorption line (BAL,
a subset of RQ), and radio-loud (RL). Radio-loud quasars are further
subdivided where suitable data are available into 
superluminal (SL), flat spectrum compact (FSC), steep spectrum compact (SSC),
and Fanaroff-Riley class 2 steep spectrum doubles (FR2). 
Further, following
the convention of Veron-Cetty and Veron (1987) and Schmidt and Green (1983), radio-quiet objects with absolute
visual magnitude fainter than $-23.0$ calculated according to their prescription
(but using our cosmological parameters, see section 5.3)
are designated as Seyfert 1 (Sy1); \nsy~objects in the \sam~%
satisfy this criterion. 
Table 1(b) lists the corresponding properties
of \nother~\other~quasars 
from our original program which were observed by \ein~or EXOSAT 
but are not included in the final \sam~%
either because of poor quality X-ray data, strong
optical variability, 
or low optical flux. 
New ground-based data are presented in this paper for these
quasars, but the data are not used in the main analysis.

Further properties of the quasars in the \sam~are listed
in Table 2, and illustrated as histograms in Fig. 2, with
the shaded regions corresponding to the radio-loud objects.
The numerical value of the radio-loudness, $R_L$, is defined
to be the logarithm of the ratio of the observed core 5 GHz flux
to the flux in the optical B band, and an object is considered
to be radio-loud if $R_L=\log(f_{5 GHz}/f_B)>1$. Note that
this definition is based on observed frame fluxes, but
corrections to the rest frame do not make an important difference
for the small redshifts encountered here.
In each case an attempt has been made to estimate a core
flux on arcsecond scales, since we are studying the properties
of the compact central source at other wavelengths and wish
to neglect the extended radio source, if any. We were unable to
find maps of the two most southern sources, and the VLA fails
to resolve the core of 3C48 from the compact steep spectrum
source in which it is embedded (Spencer et al 1989). For sources where
information on the fluxes of individual radio components was available,
we have calculated the Browne parameter $R_{CD}$ defined
as the ratio of core to extended emission at 5 GHz, and is believed
to be correlated with source orientation (Orr and Browne 1982).

Monochromatic luminosities ($\log (\nu L_\nu/ \mbox{erg/s})$) in the
visual ($L_V$ at $5400{\rm \AA}$ in the rest
frame) and X-ray ($L_X$ at 2 keV) are tabulated, as
are the bolometric luminosities derived later in this paper. 
% We
% calculate luminosities assuming a standard cosmological model with
% $H_0=50 km/s/Mpc$ and $\Omega_0=2q_0=1$.  
The absolute visual magnitude
is also given in Table 2, derived using the formulae of Veron-Cetty and Veron (1987) but
using our chosen cosmological parameters (see section 5.3).  
$\alpha_x$ is the X-ray
spectral index derived from the IPC results; the ultraviolet to X-ray
two-point spectral slope $\alpha_{ox}$ is defined from $2500{\rm \AA}$ to $2$
keV in the rest frame. 
These spectral indices are conventional energy
indices, $\alpha= -{d\log f_\nu/d\log\nu}$. 

\newpage
\section{ Data}

\subsection{Overview}

To assemble the SEDs for the \nqso~quasars required observations on 12
different telescopes, using 16 different instruments, in locations from
ground-based to space. Because of this diversity of observing techniques
it is necessary to describe each data set, and the corrections made to it,
carefully. 
\resolve{
} 



\subsection{X-ray}

The `X-ray band' covers two full decades of the spectrum,
as much as the entire UV, optical and infrared bands together.
Even though most X-ray spectra have low resolution it is still
useful to divide the band into three parts: `hard X-ray', covering
2-10 keV; `soft X-ray', covering 0.3-2 keV; and `ultra-soft X-ray'
covering 0.1-0.3 keV.
All the \nqso~objects in our sample were observed 
in 1979 or 1980
with the IPC 
instrument on {\it Einstein}.
The limited spectral resolution of this instrument allows us
to characterize the $0.1-3.5$~keV ultra-soft and soft X-ray
spectrum as a single power
law modified by foreground absorption, although the
true spectrum may be much more complex. For most of the objects
in the sample, the data and 
spectral fits are presented in WE87. 
Although the foreground Galactic absorption is now well determined for
our objects, we choose to retain the earlier fits in which the 
Galactic absorption is allowed to vary freely, since in the absence
of hard x-ray data to constrain the 1 keV slope, the presence
of an ultra-soft excess in the incident spectrum is almost equivalent to
a reduction in the absorption, and thus a better estimate
of the slope in the soft band (at greater energies than 
the excess) is obtained 
than if the absorption
is fixed at its true value (see Masnou et al 1992).

Masnou et al (1992) searched for ultra-soft 
excesses  in thirteen 
of the highest signal-to-noise observations.
Those authors re-analysed the {\it Einstein}~data 
using a two power law model and including data out to 10 keV from
the MPC instrument on {\it Einstein}, 
 revealing the presence of an
ultra-soft X-ray excess component in eight of the objects. This is indicated 
as a separate data point at 0.2 keV on our energy distribution plots (Fig. 8-54).
In these two power law fits, the presence of the hard x-ray data
means that the best estimate of the 1 keV slope is obtained by
fixing the foreground absorption at
the known Galactic (Elvis, Lockman and Wilkes 1989) value.
The object PG1211+143 was studied in a separate paper by Elvis et al (1991).
These results are summarized in Table 3(a) together with 18 previously
unpublished spectral fits. The analysis for the new fits was identical
to that of WE87.
{Contour plots for the new fits are presented in Fig. 3, and 
observational details are presented in Table 3(b); refer to WE87 for
details.}

Hard X-ray data are available for over
half the sample, from EXOSAT ME (Turner and Pounds 1989,
Comastri et al 1992) and {\it Ginga}~observations (Williams et al 1992,
Ohashi et al 1992)
and the corresponding power law fits are also listed in Table 3. 
It should be borne in mind that these power law fits are
only a parameterization; 
in higher spectral resolution observations, bright AGN
often show a complex `Compton reflection' spectrum.


\subsection{Ultraviolet}

All of the \nqso~\sam~objects have been observed with both the long and
short wavelength cameras on the {\it International
Ultraviolet Explorer}~between 1978 and 1989. 
A total of 19 exposures were made explicitly for this program,
many of which were long wavelength observations needed to fill in
pre-existing short wavelength data.
In addition, 
108 further spectra were extracted from the Regional Data
Analysis Facility archive. Both sets of data were analysed uniformly
using the GEX Gaussian extraction algorithm (Urry and Reichert 1988)
which is the most effective for our faint targets. 
Bad data points (reseau marks, cosmic ray hits and microphonic noise in
the LWR) were removed interactively.
We then averaged the ultraviolet flux within each of a set of 
wavelength bands chosen to avoid strong emission lines,
converting the somewhat noisy spectra to a small set of 
relatively well determined continuum flux estimates.

For both ultraviolet and optical data (see next section) we defined
continuum wavelength bands fixed in the
rest frame of the object, with widths $50{\rm \AA}$ wide
shortward of $1900\AA$, $100\AA$ wide shortward of $5000{\rm \AA}$ and
$200{\rm \AA}$ wide beyond that wavelength. Bands which overlapped
a region of avoidance 8000 km/s to either side of strong
broad lines or 2000 km/s to either side of strong narrow lines 
were omitted.
We avoid the lines $L\alpha \lambda1215$, $O IV \lambda 1402$,
$CIV \lambda 1549$, $CIII] \lambda1909$, $Mg II \lambda 2798$,
$Ne V \lambda3426$, $[OII] \lambda3727$, $[OIII] \lambda4959,5007$, 
$Ne II \lambda3869/3968$,
$O I \lambda 6300$, and the Balmer lines $H\alpha$ to $H\delta$.
Weak lines, the Balmer continuum, 
and the blended [FeII] lines are not avoided and so these are
included in our `continuum' fluxes. The continuum bands used
may be found in Table A1; in the published version of this
paper, Tables A1 to A5 and B1 to B47 are available only on the
AAS CD-ROM disk. Formatted, printed versions of the tables
are available from the authors on request.

Table A2 lists the IUE exposures and the observed frequencies and fluxes corresponding to each
continuum band. The fluxes are given as $\log(\nu f_\nu/\mbox{Jy Hz})$
in the observed frame, and that they are not corrected for foreground extinction.
($ 1 \mbox{Jy Hz} = 10^{-23} \mbox {erg cm$^{-2}$ s$^{-1}$}$).
The first column on each page gives the rest frame wavelengths of
each band. Then there is a header column for each object which gives the
name of the object and the logarithm of 
the corresponding observed frequencies in Hz.
This is followed by two columns for each observation giving the date
of the observation (in the first row) 
followed by the logarithmic fluxes ($\log (\nu F(\nu)/\mbox{1 Jy Hz})$)
in the first column and the uncertainties in the logarithmic fluxes 
in the second column.
The uncertainties are 1 $\sigma$ internal statistical errors derived from
the scatter in the individual data points prior to binning.

The optical spectrophotometry for the probable high redshift object PG1407+265
(see discussion below) is also included in Table A2 for convenience,
since the rest wavelengths for the observed optical spectrum lie in the ultraviolet.

\subsection{Optical}

Spectrophotometric observations for 14 objects were obtained with the
Blue ($\sim3200-6400{\rm \AA}$)
Spectrograph or the FOGS (Faint Object Grism Spectrograph, 
$\sim4500-7500{\rm \AA}$)
on the MMT. Blue spectrograph observations were made
through a 5" circular aperture at air mass below 1.4 to minimize light
lost due to atmospheric dispersion. Objects were then
re-observed at higher resolution through a 1"x3" aperture.
A nearby standard star was observed
immediately before or after the quasar observation. For FOGS 
the large aperture used was 10x20" and the
small one was a long slit 1" wide.
In all cases the large aperture
observations were used to flux calibrate the accompanying, higher S/N,
small aperture observations 
yielding spectrophotometric data with $\sim5-10{\rm \AA}$ spectral resolution.
This secondary flux calbration was made by normalising the line-free continuum
of the small aperture spectrum to that of the large aperture spectrum.
Since the amount of galaxy contamination in the lower
luminosity objects depends on aperture size, subsequent galaxy subtraction
was made with reference to the size of large aperture. The equivalent widths 
of the emission lines are those of the central point source.
They are not used in our energy distribution study and we will
discuss them no further here.
The data were reduced in the standard manner, using IRAF. To ensure
the photometric accuracy of these spectra, BVRI CCD photometric data 
were obtained on the FLWO (F. L. Whipple Observatory) 
24-inch telescope within one week of the MMT
observations.
Table 4 gives the BVRI photometry values, estimated within a 14 arcsecond 
aperture; 
the values have been corrected to the Johnson system in which Vega has magnitude $+0.03$.
In Table 4(c) we also present 
photometry obtained at Mt. Lemmon by one of us (R.C.).
This photometry is in 12 arcsecond aperture except for those
dates marked with an asterisk which are in a 17 arcsecond
aperture; see Cutri et al (1985) for observational details.
Table A3
lists the continuum fluxes observed with FOGS and MMT spectrograph,
in the same format as Table A2.
As with the IUE data, emission lines have been avoided by averaging the logarithmic fluxes in
the line-free, rest frame continuum bands listed in Table A1.


Optical spectrophotometry for the \npg~PG objects in our program 
were presented by N87. These data were already corrected
for Galactic reddening; this correction is removed here
using the same law used by those authors for consistency 
with our database (Neugebauer, G., private communication). 
This allows us, and others, to apply a uniform
Galactic dereddening to all of our optical, ultraviolet and soft X-ray data.
Since the values of  $E(B-V)$  used in N87
were not given in that paper, we tabulate them here (Table 5)
for all the objects in that paper together with the
values obtained using the prescription described below (Section 6.2).

Our figures and subsequent analysis also
include optical data from Neugebauer et al (1979), Sun and Malkan (1989),
Treves et al (1988), Sitko et al (1982), Condon et al (1981),
Adam (1978,1985), and McAlary et al (1983). These data and
other data from the literature discussed below are included
in the full energy distribution tables in Appendix B (on CD-ROM).

\subsection{Near Infrared (1-3.5$\mu$m)}

Table 6 records measurements of 
JHKL ($1.2-3.5\mu m$) photometry obtained 
explicitly for this program at the MMT and the IRTF. The table gives
the magnitudes and uncertainties,
the date of observation
and the telescope and beam size used. 
All magnitudes were derived from 
comparisons with standard stars (Tokunaga 1984, Elias et al 1982).
Magnitudes are listed on the instrumental system, i.e. without
color corrections, and correspond to a system in which Vega has
magnitude zero in all bands. The MMT observations used a filter (``N34'')
centered at 3.4~$\mu$m with spectral bandwidth 0.2~$\mu$m
instead of the standard ``L'' filter. The J filters
used at the MMT and the IRTF have an effective wavelength of 1.25
microns (Willner et al 1985). Chopper throws were
$15^{\prime\prime}$~or greater for all observations. 

We have also included infrared photometry in the figures
and analysis from 
Glass (1986), Rieke (1978), Condon et al (1981), Hyland and Allen (1982), 
Sitko et al (1982), Rudy, LeVan and Rodriguez (1982),
Ward et al (1987), and N87.



\subsection{Far Infrared ($\geq10\mu$m)}

Table 7 gives measurements
at N and Q (10, 20 $\mu$m) made at IRTF and the United Kingdom Infrared Telescope (UKIRT)
explicitly for this program.
At the IRTF, the CT1 bolometer was used with a 6" beam and an east-west
chopper throw of 30". At UKIRT the UKT8 system was used with an 8" beam
and a 20" east-west chopper throw. Magnitudes were derived as described above.

The Infrared Astronomical Satellite (IRAS, Neugebauer et al 1984) surveyed the 
sky in 1983
at far infrared wavelengths ($12-100 \mu$m). We have determined fluxes or upper 
limits
at the positions of each of our sources. Where pointed Additional
Observations (AO) were made, these were used; otherwise `lineadd' (LA) estimates
were made from the survey scans. 
This procedure allows a better estimate of the local background
and uncertainties in a particular measurement than the `coadd' maps. 
The coadded survey
maps were used to check for the presence of contaminating cirrus.
The results are listed in Table 8; they are consistent
with the results of Neugebauer et al (1986) 
and Sanders et al (1989) where we have objects in common (for the \sam,
6 objects are previously unpublished).
In some cases the $100\mu$m upper limits are rather poor because
of the presence
of cirrus in the region of the source. Upper limits ($3\sigma$) 
are listed for each source 
when no detection was made in the IRAS bands. 

\subsection{Radio and Millimetre}

We have gathered core radio fluxes at 5 GHz for the \sam~%
from the literature (Table 9). 
The PG sample study with the VLA by Kellerman et al (1989)
was given preference over other references in the calculation of radio-loudness. 
By `core' we mean the flat spectrum compact component
which appears to be physically distinct from the steep 
spectrum diffuse emission. Since this steep spectrum emission
can itself be relatively compact in angular size (e.g. 3C48),
we do not use a fixed angular size to define the core, although
in practice because detailed spectral information is not
usually available we often use flux unresolved within a 1 arcsecond
beam as our criterion. In the absence of spectral
data this is an upper limit on the flat spectrum 5 GHz core flux.
Beamsizes are given in the table.
In Table 9 we also list measurements at other frequencies
and estimates 
of the flux of any extended radio source associated with the object.

Millimetre wave data or upper limits are available for about 
half the sample (Table 10). Only III Zw 2 and 3C 273 are strong
millimetre emitters, but six other sources have weak detections.

\section{Corrections}

\subsection{Magnitude scales}

To include the optical and near infrared photometric data in
the energy distributions, we have adopted for each band an
absolute zero point appropriate
for a locally flat energy distribution.
The wavelengths for which these zero points
remain the best estimate of the energy distribution even when the
local spectrum is not flat are slightly different
from the usual nominal wavelengths of the filters,
but the uncertainties are such that we have 
retained the nominal wavelengths.
In Table 11 we tabulate for each band both the 
flux of Vega at the nominal wavelength and the
flux at the nominal wavelength of an object 
with a flat energy distribution and the same magnitude
as Vega.
We used the Hayes (1985) calibration
of Vega, extended to other wavelengths by matching to a Kurucz (1979)
theoretical model (9400K, $\log g =3.95$). We convolved the Vega energy 
distribution with atmospheric transmission curves from
Mountain et al (1985) and Manduca and Bell (1979),
filter shapes from Tokunaga (1986) for the infrared IRTF filters
and Johnson (1965) and Azusienis and Straizys (1969) for the optical bands
to obtain absolute calibrations corresponding to $\nu F(\nu)=$constant.
 The color corrections corresponding
to the difference of the Vega spectrum from a flat energy
distribution are of order 10\% for the optical bands.
The results are not very sensitive to the
details of the transmission curves; calculations using simple rectangular
bandpasses gave results which agreed to within 1\% or less except for the
U filter.
In the infrared the color corrections are smaller (of order 5\%), and the 
uncertainties
due to atmospheric transmission and the sensitivity to the shape
of the bandpass are both larger and comparable to the correction. Nevertheless,
we adopt our best estimates as described above.
The zero point of the Hayes
calibration is somewhat high relative to earlier work, and in particular
AB79 fluxes (Oke and Gunn 1983) need to be corrected upwards
by 1.5\% to agree with it, although this change is insignificant
compared to other uncertainties in the energy distributions. 
It should also be noted that since the optical photometry of Table 12
is on the Johnson system, it should be corrected for the non-zero
magnitude of Vega (+0.03 mag) before these zero points are applied.

\subsection{Extinction corrections}

Corrections for foreground (Galactic) extinction are
important in the optical, UV and soft X-ray. The 
X-ray spectral fits of Table 3 already include correction
for the line of sight absorption column as discussed above.
Separately, a single extinction correction (which may correspond
to a different hydrogen column) is
applied to the remainder of the data, using an extinction
law based on that of Savage and Mathis (1979) in the visible and ultraviolet, and Rieke
and Lebofsky (1985) in the infrared beyond $3\mu$m, Table A4.
The magnitude of the correction has
been estimated from the Galactic 
neutral hydrogen column by assuming a fixed conversion of
$N(HI)/E(B-V)=5.0\times10^{21} \mbox{cm$^2$ mag$^{-1}$}$ (Burstein and Heiles 1978). 
The Galactic HI column (Table 1) has been accurately measured with a narrow
beam and good stray radiation corrections in all but
a few cases using the 140 ft Green Bank radio telescope (Elvis, Lockman and Wilkes
1989). The value for 3C 273 was taken from Dickey, Salpeter and Terzian (1978).
In the three remaining cases it has been estimated from Heiles and Cleary (1979) or
Stark et al (1984, 1992). 

\newpage

\subsection{Cosmological model}

We have adopted a standard Friedmann-Robertson-Walker cosmological model with
 $\Omega_0 (=2q_0)=1$ and
$H_0=50 \mbox{km s$^{-1}$Mpc$^{-1}$}$. After Galactic reddening corrections were applied,
the data were blueshifted to the rest frame. Since in the rest frame we are working
with the complete energy distributions, no k-corrections and
no assumptions about the intrinsic spectrum are required.

\subsection{Host Galaxies}

Although the overall energy output of our sample objects is dominated
by the active nucleus, in the near infrared  the host galaxy
 often makes the dominant contribution. 

We have constructed a host galaxy spectral
template SED based on the Sbc galaxy model of Coleman, Wu and Weedman (1980).
We have extended this to the near infrared using  JHKL colors characteristic
of both spirals and ellipticals (Frogel 1985, de Vaucouleurs and Longo 1988).
The adopted starlight template is listed in Table A5 and shown in Fig. 4.
Fig 4. clearly shows that the  inflection in the quasar energy distribution
occurs in 
the same spectral region where the galaxy starlight dominates, emphasizing the
importance of making a correction for this starlight.
The quasar
rapidly dominates as one moves toward the optical/UV, so the difference
between spiral and elliptical templates is unimportant except in
determining the normalization. For this reason we have given preference
to infrared estimates of the galaxy luminosity over optical measurements.
We have not attempted to model the dust emission from the host galaxy,
although far-infrared observations of RSA spirals (De Jong et al 1984) and
submillimeter observations of normal galaxies
(Thronson et al 1990) indicate that the $60$ and $100$ micron contribution
of the host galaxy to the energy distribution may be significant even in the absence
of a starburst; the dashed line in Fig. 4 shows an estimate of the mean
normal spiral dust emission, but we emphasize that the dispersion in the
shape is too large to make this a useful template.
If a starburst is also present, the far infrared contribution will be even
larger, and, as Edelson (1987) and Condon and Broderick (1988) have shown,
the radio may also be contaminated.

Rather than attempt an independent fit, we have chosen to use direct
measurements of the host galaxy luminosities for the individual objects
as found in the literature. Where possible, we have used the quoted total 
apparent
magnitudes at the observed wavelengths to calculate the appropriate
normalization of the template, rather than the reported absolute magnitudes
which usually depend on the template model assumed by the authors in question.
This normalization is expressed as $L_H$, the monochromatic
luminosity at H band in the rest frame, where the host galaxy
SED peaks.
We have also attempted to estimate the model-independent half-light
radius $r_e$ for each galaxy. 

Neugebauer et al (1985) present observed frame H-band annular photometry of quasars
and estimate total magnitudes assuming that, firstly, the
radial light distribution follows a modified Hubble law
with a fixed ratio of core radius to half-light radius
as found by Kormendy (1977) and a corresponding outer cut-off;
and, secondly, that
the half light radius is related to the total absolute H magnitude
by a universal relation based on the results of Binggeli, Sandage
and Tarenghi (1984) and the assumption of a constant $B-H$ color for 
the galaxies. They then solve iteratively for the total magnitude.
We have adopted their scheme and recalculated the results for
our cosmological assumptions. We are unable to reproduce
the exact results in their paper; however,
direct measurements of the host galaxy luminosities
from CCD imaging should be available soon.
The surface brightness versus radius
relation we adopt is a modified Hubble law,
\[ \Sigma(r) = {\Sigma(0) \over (1+ {r\over 0.093r_e})^2}\]
cut-off at $r_o=5.4r_e$; experiments with a de Vaucouleurs
law gave similar results, but the de Vaucouleurs law was
a worse fit to a few of the  objects for which detailed profiles
were available - although we might expect the law to
be a better representation for those objects whose host
galaxies are elliptical.
The same scheme can be used to estimate
total magnitudes for the off-nuclear slit photometry of
Boroson, Oke and Green (1982), Boroson and Oke (1984),
and Boroson, Persson and Oke (1985); 
for these data we conducted a numerical
integration of the modified Hubble law over the area of the
slit, replacing the analytic expression possible in the annular
case. We note that our derived fractional luminosities in the slit 
ranged from 0.07 to 0.14 in reasonable agreement with the
simple assumption of 0.10 used by Boroson et al. Similarly, the
scheme can be adapted for photometry in two annuli, as in
Neizvestnii (1986) and MacKenty (1990).

For some of our objects, more detailed CCD PSF-subtracted profiles
are available. In these cases (Gehren et al 1984, Smith et al 1986,
Yee and Green 1987) we have calculated
the half-light radius and total magnitude directly from the published
profile graphs; the modified Hubble law was used to estimate
the contribution within the innermost point at which the profile
departed from the point spread function.

The total apparent magnitude and half light radius of Fairall 9 were taken 
directly from Griersmith and Visvanathan (1979).
In two cases 
 where only absolute magnitudes were given
(Q1352+183, Q2251$-$178, Hutchings, Crampton and Campbell 1984)
we took the tabulated values directly, correcting only for
cosmology.

In total, estimates of host galaxy luminosity and half-light
radius were obtained
for \nhost~of the \nqso~\sam~objects. Where no error bar
was directly available from the fitting process, we assumed
an uncertainty of 10 percent in the luminosity.
Values for the
remaining \nnohost~objects were adopted by the crude expedient of
taking the median luminosity and half-light
radius of the other hosts and using
error bars corresponding to the observed 75 percentile range;
thus $L_H=5(+3,-4)\times10^{44} \mbox{erg s$^{-1}$}$ and
$r_e=10(+3,-6) \mbox{kpc}$. The upper error bars were 
modified by the constraint that the predicted near IR flux
should not exceed that actually observed for the quasar
plus host; in only one cases, Q1244+026, was the adopted
luminosity altered to satisfy this constraint. 

Fig. 5 shows the adopted host luminosities
and half-light radii for our sample. 
The well known galaxies M31 and M87 are
marked on the diagram to orient the reader. 
Where no radial
profile was available we adopted an iterative method
similar to that of Neugebauer et al (1985) and using
the mean relationship (solid line) between half-light radius and
total luminosity found for ellipticals by Binggeli, Sandage and Tarenghi (1984)
as its starting point. 



The observed SEDs were then corrected
for host galaxy emission, using the fixed spectral host
galaxy template, the tabulated normalization and errors, and
the modified Hubble law, which was used to determine the fraction
of the host flux within the aperture separately for each data point.
Uncertainties in the half-light radius were, however, ignored
in the calculation of the final errors.

Table 12 gives the adopted normalizations at rest frame H for the
host galaxy luminosities; most are a few times $10^{44} \mbox{ erg s$^{-1}$}$. 
We also present the adopted half-light radii in kpc. 
The size of the correction at H is shown as a function
of absolute magnitude in Fig. 6.
{Corrections can be important up to $M_V=-25.5$ or more (e.g. 3C 48),
while quasars as faint as $M_V=-23.5$ can show negligible host
galaxy contamination (e.g. Kaz 102). }



\subsection{Variability and averaging}

A potentially severe limitation on our dataset is that the observations are typically
not simultaneous, although the
optical and ground based IR data were generally obtained within about one
month. This problem is worst in the ultraviolet since the amount
of variability increases with frequency throughout the UVOIR region (Cutri
et al 1985). Edelson, Krolik and Pike (1990) reported that the degree of 
variability
was smaller in the higher luminosity sources that form the
bulk of the present sample, than in low luminosity Seyferts. However,
the variations are still enough to contribute to the scatter in our ultraviolet
energy distributions.

For about one third of the objects we have observations 
at two epochs (occasionally more) in a given waveband, so we can make a
crude estimate of the degree of variability. In Table 13 we list
the observed range of variability, $F_{max}/F_{min}$,
and the associated timescale, at rest wavelengths in the 
near infrared, optical, mid ultraviolet and far ultraviolet.
It can be seen that in the optical and infrared variability is not a serious
problem for these `normal' quasars, but that in the ultraviolet
the variability is significant on timescales of a few years (median
timescale sampled is 4 years), although
typically (13 out of 18 cases) it is less than a factor of two.
These results indicate somewhat less variability than found by
Kinney et al (1991), and this is probably an artifact of the small
number of observations of our objects (more observations would
tend to increase the observed range of variability).

To generate a single mean energy distribution for each quasar, we
have taken an average (in $\log \nu F(\nu)$) of all the data in each frequency bin.
However, for the IUE data, because of the increased problem of variability
and also the widely different S/N among observations,
we have been selective in the exposures from Table A2
we chose to include in the average. Specifically, where simultaneous
data from the long (LWP/LWR) and short (SWP) wavelength
cameras were available, we have included them and excluded
`orphan' LWP/LWR or SWP exposures. This avoids spurious steps
in the UV data due to variability. Objects where `orphans'
were excluded are Q0007+106, Q1100+772,
Q1146-037, Q1202+281, Q1407+265, and Q1613+658.
Further, when one exposure had significantly lower signal-to-noise than the others,
it was omitted (Q1545+210, Q1613+658, and  Q1721+343).

\subsection{PG 1407+265}

The redshift of PG 1407+265 (= 2E 3196) is uncertain; the object has
very weak emission lines, the only certain feature coming at $5500{\rm \AA}$. An
identification of this as Mg II and the weak presence of a feature at CIII] 
led Schmidt and Green (1983) to
propose the redshift as z=0.944.  The near infrared to ultraviolet spectrum
of PG1407+265 is presented in Fig. 7. 
The absence of a clear Lyman
alpha line in the IUE spectrum caused us to consider lower
redshift identifications, including blueshifts, but no choice of redshift allows
normal quasar line ratios. We have adopted the high redshift given by Schmidt and Green
because of the position of a prominent continuum feature, 
namely the near infrared inflection. The rest wavelength
of this feature lies between 1.0 and 1.5 microns for all our other objects,
and in this object occurs at an observed wavelength between 2.0 and 2.4 microns,
which lends support to the high redshift estimate.

We continue to adopt the Schmidt and Green value despite the absence of
definitely observed hydrogen lines.  We note that the broad width of the one
line clearly detected, the overall continuum shape, the lack of strong optical
variability and the X-ray to optical flux ratio indicate that the object
is indeed a quasar rather than, for instance, a star, a BL Lac object,
or some more exotic object.
Note the good
agreement in flux level at the overlap between the two optical spectra
and the smooth continuation of the spectrum made by the infrared
photometry, implying a lack of variability at longer wavelengths. 
Comparison with the data in Neugebauer et al (1987) indicates
that the optical has varied by less than 10 per cent over an 8 year
interval.
A 1986 LWP spectrum shows a lower flux level than the other ultraviolet
data, implying significant variability in the UV (at least 30\%).
These variability properties are all consistent with normal
quasar behaviour.
  
\newpage
\section{Properties of the energy distributions}

In Figs. 8-19 and 20-35 we present two energy distributions for each object
in the \sam, an overall view
and a closeup of the infrared to ultraviolet region. The overall view
covers a fixed flux range of ten decades and illustrates the radio-loudness
and X-ray properties of each quasar. The closeup view covers two decades
in flux and illustrates the near infrared inflection, and the strength of the
blue bump. For objects in which host galaxy subtraction makes
a significant difference ($>5$\%) to the energy distribution, the starlight
subtracted energy distribution is also shown.

\subsection{Luminosities: Bolometric and individual bands}


To characterize the large scale distribution of the energy output of the quasars, we
calculate integral luminosites in a set of broad bands. The integrals
are calculated by running a simple linear interpolation through
the data points in $\log \nu L_\nu$ space, i.e. connecting the individual
points with a power law. The errors indicated in the tables are estimated
by performing the integrals using the one sigma high and one sigma low
flux values instead of the nominal values. For upper limits 
we interpolate between detections on either
side. The lower of the interpolated value and the upper limit
is used as the nominal flux estimate, but the errors are estimated
using zero as the lower error bar and the upper limit as the
upper error bar. 

The logarithms of the 
calculated integral luminosities 
in units of \mbox{erg s$^{-1}$} are tabulated in Tables 14 to 16.

\begin{enumerate}

\item{{\em Bolometric}}
The bolometric luminosity is typically well defined except for
contributions from two regions of the spectrum: the mostly unobservable EUV region 
and the as yet unobserved hard X-ray and gamma-ray region.
Indications are strong that the gap in the 0.1-10 mm wavelength
range is energetically
negligible.
As a first crude estimate of the EUV luminosity, 
we simply make a linear interpolation  (in the log-log space
of Figs. 8-19, i.e. a power law interpolation in flux space) 
between the ends of the IUE and
{\it Einstein}~ranges. We note, however, that in contrast to 
the claim of Padovani and Rafanelli (1988), the lack of EUV data
does introduce a major uncertainty into the final bolometric
luminosity. An upper limit to the physically reasonable EUV luminosity
can be made by finding the maximum blackbody curve which does not
exceed the observed data; however this limit is not strong, as the
luminosity implied is typically 10 to 100 times the luminosity
observed in the rest of the spectrum. 

We perforce neglect the unknown luminosity above
10 keV ($10^{18.4}$Hz). Compton Observatory observations
will provide some constraints in this region.

The number of ionizing photons is somewhat better determined than
the luminosity, since the photon spectrum is rapidly falling
in the ultraviolet. However our estimate of this quantity is
still very speculative since it is based on our linear interpolation
across the EUV gap. We tabulate the total estimated ionizing photon
rate multiplied by 1 Rydberg, $N_{Ion}R$, to give the quantity the same units
as the luminosities for easy comparison.
Since $1 \mbox{Ryd}= R = 2.18\times10^{-11} \mbox{erg},$ if $N_{Ion}R=10^{44}N_{44}
\mbox{erg s$^{-1}$}$ then $N_{Ion}=4.6\times 10^{54}N_{44} \mbox{photons s$^{-1}$}.$
The mean ionizing photon energy
in Rydbergs is then simply $L_{Ion}/(N_{Ion}R).$ 

\item{\em UVOIR}
The three decades between 100$\mu$m and 0.1$\mu$m (the ultraviolet/\- optical/\-infrared
or `UVOIR' region) are relatively well sampled, so the corresponding
UVOIR luminosity can be calculated much more accurately. This luminosity 
accounts for most of that which is directly observed, and so is a useful
fiducial luminosity for a quasar.

\item{\em Decades}
We also tabulate
the luminosity in individual decades across the electromagnetic spectrum.
Outside the range $1-0.1\mu$m, these are often estimates
from only one or two points and the errors are correspondingly
large (typically 25 percent in the far IR).

\item{\em Octaves}
In order to describe the shape of the ultraviolet bump component,
we define a set of narrower octave wide bands.
The four bands we call IR ($0.8-1.6\mu$m), 
VIS ($4000-8000{\rm \AA}$), NUV ($2000-4000{\rm \AA}$)
and UV ($1000-2000{\rm \AA}$). The IR band gives the luminosity just shortward
of the near infrared inflection, where the bump starts.
The UV band measures the luminosity in the bluest observed
part of the bump, while the VIS band samples the early part of the bump's rise.
The NUV band covers the near ultraviolet region of the spectrum dominated by the `small bump'
of blended Fe II and Balmer continuum emission (Wills, Netzer and Wills 1985).
Further discussion and interpretation of the octave colors of our sample may be found
in Kuhn et al (in preparation).

\end{enumerate}

\newpage
\subsection{The Mean and Dispersion of the SEDs}

We have also derived mean bolometric corrections and mean
energy distributions for the sample.
We have excluded four
objects (Q0923+129, Q1244+026, Q1351+695, and Q2209+184) from
these calculations
because of the large
uncertainties in the starlight subtraction for those objects.
In Table 17, 
the median, mean and standard deviation of the bolometric
corrections at a given band
are given, followed by the minimum and maximum values
found in the sample. Errors in the determination of
individual energy distributions have been ignored for
the purposes of this table. 
To estimate bolometric luminosity from a rest frame
luminosity at B, the value of $\nu L_\nu$ at B may be multiplied
by the  median derived visual bolometric correction in Table 17, approximately
a factor of 12. This is somewhat smaller than the value
of 16.5 derived by Sanders et al (1989) in their study, and 
the range of actual values of the correction covers a factor of 5. 
The median value
of $N_{Ion}R/L_{Bol}$ found in our sample is 0.12. This means that
for a quasar of bolometric luminosity of  $10^{44} \mbox{erg s$^{-1}$}$ 
the corresponding median rate of emission of ionizing photons is
$N_{Ion}=5.5\times10^{53} \mbox{photons s$^{-1}$}.$ The median value of
the mean ionizing photon energy is found to be 2.8 Rydberg.

In figures 36 and 37 we present the quasar mean energy distribution (MED).
This distribution is obtained by normalizing the individual starlight-subtracted 
energy
distributions and smoothing them
with a $\Delta\log(\nu)=0.2$ boxcar. The mean of $\log(\nu L\nu)$ 
is then calculated
using the Kaplan-Meier estimator (Feigelson and Nelson 1985).
Figure 36 shows the MED for radio-quiet and radio-loud quasars
separately, over the full frequency range we have studied. Note 
that the different X-ray slopes for radio-loud and radio-quiet
quasars show up clearly, but that the MED for radio-loud
and radio-quiet quasars is indistinguishable in the UVOIR region.
We have normalized the MEDs 
at $\munorm~\mu\mbox{m}$, the approximate mean location
of the near infrared inflection point.
While the MED is a good description of the typical quasar, 
the full range of SEDs contains considerable variety.
Figure 37(a) shows the MED for the combined radio-loud and radio-quiet
sample in the UVOIR region, together with contours illustrating
the dispersion of the actual energy distributions about the median.
The Kaplan-Meier median is in very close agreement with the mean
calculated in Fig. 36.
Indicated are 68, 90, and 100 percentile contours on
each side of the median, calculated
using the smoothed Kaplan-Meier estimate of the survival function (Feigelson
and Nelson 1985, Miller 1981). The narrowness at \munorm~microns
is an artifact of the normalization there; Figure 37(b) shows
the results obtained when the energy distributions are normalized
by their total UVOIR (100 micron -- 1000 $\AA$) luminosity.

The 68 percentile distribution is within a factor of 2-3 of the
mean throughout, so the mean is revealing something about the
nature of quasars. However, the 90 percentile width is a factor of 20,
and the extremes are a factor of 30 in the UV and 60 in the far
IR. The distributions are
broader than would be expected from a Gaussian distribution.
The MED and its dispersion will be discussed in more detail in
a future paper. Here we note that the large dispersion of
shapes in individual objects means that the MED should be used
only with caution, and that the variety of shapes should contain
information about the physics of quasars.
Our MED is in 
good agreement with that of Sanders et al (1989) except
in the far infrared at 60 microns and beyond, where the
inclusion of upper limits lowers the estimate of the mean. 
Our results for radio-quiet quasars are compared with the corresponding
Sanders et al results in Fig. 38.

\section{Conclusions}

We have collected a set of data useful for many investigations.
In particular, these data will serve as a benchmark against
which to compare the energy distributions of high redshift 
quasars.  The data are available in FITS BINTABLE format by anonymous ftp from
sao-ftp.harvard.edu in directory `pub/jcm/qed'. 
The uncorrected, observed frame data are tabulated in Appendix B.


\acknowledgements

This work was carried out as part of NASA Astrophysics Data Program
grant NAG8-689 and Long Term Space Astrophysics
Research Program grant NAGW-2201, NASA HEAO contract NAS 8-30751,
and NASA IUE grants NAG5-87, NAG5-37.
We acknowledge useful discussions with 
%[include AAS membership list here, esp.]
Richard Barvainis, Ski Antonucci, Nat Carleton, Walter Rice,
Olga Kuhn, Aneta Siemiginowska, Bozena Czerny, and Diana Worrall.
Data for this paper were obtained from the Einstein Data Bank, the 
IUE Reduction and Data Analysis Facility, and the IRAS data bank at IPAC,
and we thank the staff at those facilities for their support.
JCM thanks the National Research Council for support as a NAS/NRC
Associate during part of this project. BJW acknowledges
support from the US Rosat Science Data Center.
\newpage

\begin{references}

\reference
\phantom{reference}

\reference
Adam, G. 1978, A\&AS {\xem 31}, 151
 
\reference
Adam, G. 1985, A\&AS {\xem 61}, 225
 
 
\reference
Antonucci, R. R. J., and Barvainis, R. 1988, ApJ {\xem 325}, L21

\reference
Antonucci, R. R. J., Barvainis, R., and Alloin, D. 1990, ApJ 
{\xem 353}, 416

\reference
Arnaud, K.A., et al. 1985, MNRAS {\xem 217}, 105

\reference
Azusienis, A., and Straizys, V. 1969, AZh {\xem 46}, 2, 402

\reference
Barvainis, R., Antonucci, R. R. J. 1989, ApJS 70, 257

\reference
Bezler M., Kendziora E., Stubert R., Hasinger G., Pietsch W., Reppin C., 
Truemper J., Voges W. 1984, A\&A {\xem 136}, 351
 
\reference
Binggeli, B., Sandage, A., and Tarenghi, M. 1984, AJ, {\xem 89}, 64

\reference
Bolton, J.G., and Butler, P.W. 1975, Australian J. Phys. Suppl., {\xem 34}, 33

\reference
Bolton, J. G., Wall, J. V., Shimmins, A. J. 1971, Australian J. Phys. {\xem 24}, 889

\reference
Boroson, T.A., and Oke, J.B. 1984, ApJ {\xem 281}, 535

\reference
Boroson, T.A., Oke, J.B., and Green, R.F. 1982, ApJ {\xem 263}, 32

\reference
Boroson, T.A., Persson, S.E, and Oke, J.B. 1985, ApJ {\xem 293}, 120


\reference
Burstein, D., and Heiles, C. 1978, ApJ, {\xem 225}, 40

\reference
Carleton N.P.,Elvis M., Fabbiano G., Lawrence A., Ward M.J. and
Willner S.P. 1987, ApJ {\xem 318}, 595

\reference
Chini, R., Kreysa, E. and Biermann, P.L. 1989,
A\&A {\xem 219}, 87
 
\reference
Clegg, P. et al. 1983, ApJ {\xem 273}, 58

\reference
Coleman, G.D.,  Wu, C.C.,  and Weedman, D.W. 1980, ApJS {\xem 43}, 393
 
\reference
Comastri, A., Setti, G., Zamorani, G., Elvis, M., Wilkes, B.J., 
McDowell, J.C., and Giommi, P. 1992, ApJ {\xem 384}, 62

\reference
Condon, J. J., and Broderick, J. J. 1988, AJ {\xem 96}, 30

\reference
Condon J. J., O'Dell S. L., Puschell J. J. and Stein W. 1981, 
ApJ {\xem 246}, 624
 
\reference
Cutri R. M., Wisneiewski W. Z., Rieke G. H. and Lebofsky M. J. 1985, ApJ
{\xem 296}, 423
 
\reference
Czerny, B. and Elvis, M. 1987, ApJ {\xem 321}, 305
 
\reference
de Jong, et. al. 1984, ApJ {\xem 278}, L67

\reference
de Vaucouleurs, A., and Longo, G. 1988, {\it Catalog of Visual
and Infrared Photometry of Galaxies from $0.5\mu$m to $10 \mu$m (1961-1985),}
Univ. Texas Mono. Astr., No. 5.

\reference
Della Ceca, R., Palumbo, G. G. C., Persic, M., Boldt, E. A., DeZotti, G., and
Marshall, F. E. 1990, ApJS {\xem 72}, 471

\reference
Dickey J. M., Salpeter, E.,  and Terzian, Y. 1978, ApJS {\xem 36},77
 
\reference
Edelson, R. A. 1987, ApJ {\xem 313}, 651

\reference
Edelson, R. A., Krolik, J. H. and Pike, G. F. 1990, ApJ {\xem 359}, 86

\reference
Edelson, R. A. and Malkan, M. A. 1986, ApJ {\xem 308}, 59

\reference
Edelson, R. A., Malkan, M. A., and Rieke, G. H. 1987, ApJ {\xem 321}, 233

\reference
Ekers, J. A. 1969, Australian  J. Phys. Suppl., {\xem 7}, 1

\reference
Elias, J. H. Frogel, J. A., Matthews, K., and Neugebauer, G. 1982,
AJ {\xem 87}, 1029

\reference 
Elsmore, B., and Ryle, M., 1976, MNRAS 174, 411

\reference 
Elvis M., Green R. F., Bechtold J., Schmidt M., Neugebauer G., Soifer
B. T., Matthews K. and Fabbiano G., 1986, ApJ, {\xem 310}, 291

 
\reference
Elvis, M., Lockman, F.,  and Wilkes, B.J. 1989, AJ {\xem 97}, 777
 
\reference
Elvis, M., Giommi, P., Wilkes, B.J., and McDowell, J. C. 1991,
ApJ, {\xem 378}, 537

\reference
Engargiola, G., Harper, D.A., Elvis, M., and Willner, S. P. 1988,
ApJ, {\xem 332}, L19

\reference
Ennis, D. J., Neugebauer, G.,  and Werner, M. 1982, ApJ {\xem 262}, 460
 
\reference
Feigelson, E. D., Isobe, T. and Kembhavi, A. 1984, AJ {\xem 89}, 1464

\reference
Feigelson, E. D., and Nelson, P. I. 1985, ApJ 293, 192

\reference
Frogel, J. 1985, ApJ {\xem 298}, 528

\reference
Gehren, T., Fried, J., Wehinger, P.A., and Wyckoff, S., 1984,
ApJ, {\xem 278}, 11

\reference
Giacconi, R., et al. 1979, ApJ {\xem 230}, 540

\reference 
Glass, I. S. 1986, MNRAS {\xem 219}, 5P

\reference
Gorenstein, P., Harnden, R.F., and Fabricant, D. 1981,
IEEE Trans. Nucl. Sci., NS-28, 869

\reference
Gower, A.C. and Hutchings, J.B. 1984. AJ {\xem 89}, 1658

\reference
Gregory, P. C., and Condon, J. J. 1991, ApJS {\xem 75}, 1011

\reference
Griersmith, D., and Visvanathan, N. 1979, A\&A
{\xem 79}, 329
 
\reference
Harris, D.E., et al. 1991, {\it The Einstein Observatory
Catalog of IPC X-ray Sources}, SAO HEAD CD-ROM Series 1 (Einstein),
Set 4, Disk No. 7.
 
\reference
Hayes, D. S., 1985, in IAU Symposium 111, eds. Hayes, D.S. et al., 225.

 
\reference
Heiles. C.,  and Cleary, M.N. 1979, Australian J. Phys. Suppl., {\xem 47}, 1

\reference
Hintzen, P., Ulvestad, J., and Owen, F. 1983, AJ {\xem 88}, 709

\reference
Hutchings, J.B., Crampton, D., and Campbell, B. 1984, ApJ {\xem 280}, 41

\reference
Hutchings, J.B.,  and Gower, A.C.,  1985, AJ {\xem 90}, 405.

\reference
Hutchings, J.B., Janson, T., and Neff, S. G., 1989, ApJ {\xem 342}, 660.

\reference
Hutchings, J.B., Johnson, I., and Pyke, R. 1988, ApJS 
{\xem 66}, 361
 
\reference
Hutchings, J.B., Price R., and Gower A.C. 1988, ApJ 329, 122

\reference
Hyland, A.R.,  and Allen, D.A. 1982, MNRAS {\xem 199}, 943
 
\reference
Jagers, W. J., van Breugel, W. J. M., Miley, G. K., Schillizzi, R. T., and
Conway, R. G. 1982, A\&A {\xem 105}, 279

\reference
Johnson, H. L. 1965, ApJ {\xem 141}, 923

\reference
Kellermann, K. I., Sramek, R., Shaffer, D., Green, R., and Schmidt, M.
1989, AJ {\xem 98}, 1195

\reference
Kinney, A, Bohlin, R. C., Blades, J. C., and York, D. G. 1991,
ApJS {\xem 75}, 645

\reference
Kormendy, J. 1977, ApJ {\xem 218}, 333

\noindent
Kriss, G. 1988. Ap.J,  {\xem 324}, 809.    
 
\reference
Kurucz, R. L. 1979, ApJS {\xem 40}, 1

\reference
Landau R., Epstein, E. E., and Rather, J. D. G. 1980, AJ
{\xem 85,} 363
 
\noindent
Landau R., et al, 1986, Ap.J., {\xem 308}, 78.
 
\reference
Lonsdale C., and Morison I., 1983, MNRAS, 203, 833

\reference
Low, F. J., Huchra, J. P., Kleinmann, S. G., and Cutri, R. M. 1988,
ApJ {\xem 327}, L41

\reference
MacKenty, J.W. 1990, ApJS {\xem 72}, 231

\reference
Malkan M. A. 1983, ApJ {\xem 268}, 582
 
\reference
Malkan M. A. and Sargent W. L. W. 1982, ApJ {\xem 254}, 22
 
\reference
Manduca, A. and Bell, R. A. 1979, PASP 91, 848.

\reference
Masnou J.-L., Wilkes, B. J., Elvis, M., McDowell, J. C., and Arnaud, K. A.
1992, A\&A {\xem 253}, 35

\reference
McAlary, C. W., McLaren, R. A., McGonegal, R. J., and Maza, J. 1983,
 ApJS {\xem 52}, 341
 

\reference
Meurs, E. J. A., and Wilson, A.S. 1981, A\&AS {\xem 45}, 99

\reference
Miley, G. K., and Hartsuijker, A. P. 1978, A\&AS
{\xem 34}, 129

\reference
Miller, R.G., Jr., 1981, Survival Analysis (Wiley: New York).

\reference
Mountain, C.M., Leggett, S. K., Selby, M. J. and
Zadrozny, A. 1985, A\&A 150, 281

\reference
Mushotzky R.F. 1984, Advances in Space Research, {\xem 3}, no. 10-13, 312
 
\reference
Neizvestnii, S.I. 1986, Soobshcheniya Spetsial'noi Astrofisicheskoi
Observatorii, 51, 5.

\reference
Neugebauer, G., Oke, J. B., Becklin, E. E., and Mathews, K. 1979,
ApJ {\xem 230}, 79
 
\reference
Neugebauer, G., et. al. 1984, ApJ {\xem 278}, L1

\reference
Neugebauer, G., Matthews, K., Soifer, B.T., and Elias, J.H. 1985,
ApJ {\xem 298}, 275

\reference
Neugebauer, G., Miley, G.K., Soifer, B.T., and Clegg, P.E.
1986, ApJ {\xem 308}, 815

\reference
Neugebauer G., Green R.F., Matthews K., Schmidt M., Soifer B.T. and
Bennett J. 1987, ApJS, {\xem 63}, 615 (N87).
 
\reference
Ohashi, T, et al. 1992, ApJ {\xem 398}, 87

\reference
Oke, J., and Gunn, J. 1983, ApJ {\xem 266}, 713

\reference
Orr, M.J.L., and Browne, I.W.A. 1982, MNRAS {\xem 200}, 1067

 
\reference
Owen, F., Porcas, R. W., and Neff, S. G. 1978, AJ {\xem 83}, 1009

\reference
Owen, F., Porcas, R. W., Mufson, S. L., and Moffett, T. J. 1978,
AJ, {\xem 83}, 685
 
\reference
Owen, F., and Puschell, J. 1982,  AJ {\xem 87}, 595.

\reference
Padovani, P., and Rafanelli, P. 1988, A\&A {\xem 205}, 53.

\reference
Perley, R. A. 1982, AJ {\xem 87}, 859
 
\reference
Pooley, G. G.,  and Henbest, S. N. 1974, MNRAS {\xem 169}, 477
 
\reference
Preston, R.A., et al. 1985, AJ {\xem 90}, 1599

\reference
Price, R.M., and Milne, D.K. 1965, Australian J. Phys, {\xem 18}, 329
 
\reference
Rieke, G. H. 1978, ApJ {\xem 226}, 550
 
\reference
Rieke, G. H. and Lebofsky, M.J. 1985, ApJ {\xem 288}, 618
 
\reference
Robson, E. I., Gear, W. K., Smith, M. G., Ade, P. A. R., and Nolt, I. G. 1985,
MNRAS {\xem 213}, 355
 
\reference
Rudnick, L., Sitko, M. L., and Stein, W. A. 1984, AJ {\xem 89}, 753
 
\reference
Rudy, R. J., LeVan, P. D. and Rodriguez-Espinoza, J. M. 1982, AJ {\xem 87}, 598
 
\reference
Sandage A. 1965, ApJ {\xem 141}, 1560

\reference 
Sanders, D. B., et al. 1989, ApJ {\xem 347}, 29

\reference
Savage, B. D., and Mathis, J.S. 1979, ARA\&A
{\xem 17}, 73
 
\reference
Saxton, R.D., et al. 1993, MNRAS, 262, 63

\reference
Schmidt M. and Green R.F. 1983, ApJ {\xem 269}, 352
 
\reference
Shields G.A. 1978, Nature, {\xem 272}, 706.

\reference
Shimmins, A.J., and Bolton, J.G. 1972a, Australian J. Phys. Ap. Suppl, 
{\xem 23}, 1.
 
\reference
Shimmins, A.J., and Bolton, J.G. 1972b, Australian J. Phys. Ap. Suppl, 
{\xem 26}, 1.
 
\reference
Shimmins, A.J., and Bolton, J.G. 1981, Australian J. Phys. 34, 471.
 
\reference
Shimmins, A. J., Day, G. A., Ekers, R. D., and Cole, D. J. 1966,
Australian J. Phys. {\xem 19}, 837

\reference
Sitko, M.L., Stein, W.A., Zhang, Y-X, and Wisniewski, W.Z. 1982,
ApJ {\xem 259}, 486
 
\reference
Smith, E., Heckman, T.M., Bothun, G.D., Romanishin, W., and 
Balick, B. 1986, ApJ {\xem 306}, 64

\reference
Spencer, R.E., McDowell, J.C., Charlesworth, M.,
Fanti, C., Parma, P., and Peacock, J.A.
1989, MNRAS {\xem 240}, 657
    
\reference
Stark, A.A., Heiles, C., Bally, J., and Linke, R., 1984,
privately distributed magnetic tape.

\reference
Stark, A. A., Gammie, C. F., Wilson, R. W., Bally, J.,
Linke, R. A., Heiles, C., and Hurwitz, M. 1992,
ApJS 79, 77

 
\reference
Steppe, H., Salter, C.J., Chini, R., Kreysa, E., Brunswig, W., 
and Lobato Perez, J. 1988, A\&AS {\xem 75}, 317


\reference
Sun, W.-H., and Malkan, M. 1989, ApJ {\xem 346}, 68

\reference
Swarup, G., Sinha, R.P., and Hilldrup, K. 1984, MNRAS {\xem 208},
813

\reference
Tananbaum, H., Avni, Y., Green, R.F., Schmidt, M., and Zamorani, G.
1986, ApJ {\xem 305}, 57

\reference
Thronson, H.A., Jr., Hunter, D.A., Casey, S., Engargiola, G., and
Harper, D.A., 1990, in {\it 2nd Wyoming Conference: The Interstellar
Medium in External Galaxies}, ed. D.J. Hollenback and H.A. Thronson, Jr.,
NASA CP-3084, p. 116.

\reference
Tokunaga, A. 1984, AJ {\xem 89}, 172

\reference
Tokunaga, A. 1986, {\it IRTF Photometry Manual.}

\reference
Treves, A., Bouchet, P., Chiapetti, L, Ciapi, A., Falomo, R., 
Maraschi, L., and Tanzi, E.G. 1988, ApJ {\xem 330}, 178.

\reference
Turner, T.J. and Pounds, K.A. 1988, MNRAS {\xem 232}, 463.

\reference
Turner, T.J. and Pounds, K.A. 1989, MNRAS {\xem 240}, 833.

\reference
Unger, S.W., Lawrence, A., Wilson, A.S., Elvis, M., and Wright, A.E.  1987,
MNRAS, {\xem 228}, 521.
 
\reference
Urry, C.M.,  and Reichert, G. 1988, NASA IUE Newsletter no.34, p.95.

\reference
Van Breugel W., Miley G., and Heckman T., 1984, 
AJ 89, 5

\reference
Veron-Cetty, M.P., and Veron, P. 1987, {\it  ESO Scientific Report,}
No. 5.

\reference
Ward, M., Elvis, M., Fabbiano, G., Carleton, N.P., Willner, S.P., and
Lawrence, A. 1987, ApJ {\xem 315}, 74.

\reference
White, R.L., and Becker, R.H. 1992, ApJS {\xem 79}, 331.

\reference
Wilkes, B.J., and Elvis, M. 1987, ApJ {\xem 323}, 243 (WE87).

\reference
Williams, O. R., et al. 1992, ApJ {\xem 389}, 157.
 
\reference
Willner, S. P., Elvis, M., Fabbiano, G., Lawrence, A., and Ward, M. J.
1985, ApJ 299, 443

\reference
Wills, B.J., 1975,  Australian J. Phys. Ap. Suppl, 
{\xem 38}, 1.

\reference
Wills B.J., Netzer H. and Wills D. 1985, ApJ, 288, 94.
 
\reference
Wills, D. 1979, ApJS {\xem 39}. 291.


\reference
Wright, A.E., Wark, R.M., Troup, E., Otrupcek, R., Jennings, D,
Hunt, A., and Cooke, D.J. 1991, MNRAS {\xem 251}, 330.

\reference
Yee, H.K.C., and Green, R.F. 1987, AJ {\xem 94}, 618.

\reference
Zamorani, G., et al. 1981, ApJ {\xem 245}, 357.

\end{references}
\newpage 

\centerline{\bf{Figure captions}}

\vskip 0.2in

\noindent
Fig. 1. Examples of radio-loud (4C 34.47, top) and radio-quiet (Mkn 586,
bottom) quasar energy distributions, illustrating the main continuum
features. The energy distributions show the logarithm of the
energy per unit logarithmic frequency interval, in the rest frame.

\vskip 0.2in

\noindent
Fig. 2. Histograms of sample properties. Shaded bins
correspond to radio-loud objects, $R_L>1.$ The absolute magnitude
is calculated using the corrections of Veron and Veron (1987),
but with our value of $\Omega_0$. The bolometric luminosity
is derived from the observed energy distributions as described in the text.
The monochromatic X-ray luminosity is $\nu L(\nu)$ at 2 keV in the rest frame,
estimated from the IPC spectral fits. $R_{CD}$, the ratio of radio
core to extended luminosity (Orr and Browne 1982), is an indicator
of the source orientation; it was only possible to estimate this
for a subset of the sources. 

\vvs
Fig 3. Contour plots of $\chi^2$ goodness of fit as
a function of fitted spectral index and foreground hydrogen
column density. For details, refer to Wilkes and Elvis (1987).
The vertical line indicates the actual galactic hydrogen column.
The contours correspond to an increase of $\chi^2$ above
the minimum of 2.30, 4.61 and 9.21.

\vvs
Fig 4. Starlight template used for host galaxy subtraction.
Template is shown with energy distribution of PG 1426+015 
for comparison, with relative
normalization appropriate for the limit of large aperture.
Dashed lines illustrate mean IR and X-ray
colors of galaxy, not used in subtraction.


\vvs
Fig 5. Adopted host galaxy luminosities and radii.
The solid curve indicates the Bingelli et al (1984)
relation. Points marked by open circles are constrained
to lie on the curve, while radii for other points are derived
using radial profile information as described in the text. The median
host galaxy normalization used where specific information
is lacking is indicated by the large solid symbol and
represents the median of the other points. The well known
galaxies M31 and M87 are also plotted on the figure 
for comparison.


\vvs
Fig. 6 Host galaxy flux corrections at rest frame H (percent)
as a function of absolute visual magnitude. Solid circles
indicate corrections made using data on the specific host galaxy,
while open circles indicate use of the median normalization.

\vskip 0.2in
\noindent
Fig 7. The observed energy distribution of PG1407+265 from the near
infrared to the far ultraviolet. The expected positions of prominent
quasar emission lines are indicated for an assumed redshift $z=0.94$.


\vskip 0.2in
\noindent
Figs. 8-19. Rest frame, dereddened continuum 
energy distributions of the quasar sample.
For each object, the panel shows the overall radio to X-ray
energy distribution.

Figs. 20-35. Rest frame, dereddened continuum
energy distributions of the quasar sample, 
showing the details of the
UVOIR ($100 \mu$m to $1000{\rm \AA}$) region. 
When two panels are present for an object, the upper panel
shows the data before host galaxy subtraction, and the lower
panel shows the same region after
host galaxy subtraction (see text).

Fig. 36. The mean quasar energy distribution, normalized
at \munorm~microns, for radio-loud (dashed line) 
and radio-quiet (solid line) quasars. Spectral regions where
little or no data are available are omitted.
The radio-loud distribution has a rising x-ray spectrum in this plot,
while the radio-quiet x-ray spectrum is horizontal.

Fig. 37. (a) The median quasar energy distribution in the
UVOIR range, normalized
at \munorm~microns, and the 
68, 90, and 100 (dashed) Kaplan-Meier percentile envelopes, showing the large
dispersion from the mean in the far infrared and ultraviolet.
(b) The median and percentiles when the energy distributions are normalized
by their total 100 micron to 1000$\AA$ luminosity rather than the
monochromatic \munorm~micron luminosity.

Fig. 38. Comparison of mean energy distributions for radio-quiet
quasars. Solid line: UVSX sample using Kaplan-Meier estimator (this paper). Dashed line: 
UVSX sample with conventional mean, excluding upper limits. Open circles: PG quasars from 
Sanders et al (1989), using conventional mean.
\end{document}

\newpage
\centerline{\bf{Table captions}}

\vvs
Table 1. Quasars in the Atlas. (a) The \sam;
(b) other quasars for which new observations
are tabulated. The $N_H$ value
is in units of $10^{20}$ cm$^{-2}$. The `Ref' column indicates
references for the $N_H$ value: (1) Elvis, Lockman
and Wilkes (1989); (2) Heiles and Cleary (1989); 
(3) Dickey, Salpeter and Terzian (1978); (4) Stark et al (1984,1992).
The classes of object based on optical luminosity
and radio spectrum and morphology are: RQ (Radio-quiet); Sy1 (Low
luminosity radio-quiet, or Seyfert 1); BAL (Radio-quiet
with broad absorption lines); RL (Radio-loud); SL (Radio-loud
superluminal); FSC (Flat spectrum compact radio-loud);
SSC (Steep spectrum compact radio-loud); and
FR2 (Radio-loud Faranoff-Riley class 2 steep spectrum doubles).

\vvs
Table 2. \sam~quasars: further properties.
{
%\par\noindent 
(1) Monochromatic luminosity (erg cm$^{-2}$ s$^{-1}$) at 5400${\rm \AA}$;
%\par\noindent 
(2) Monochromatic luminosity (erg cm$^{-2}$ s$^{-1}$) at 1 keV;
%\par\noindent 
(3) Bolometric luminosity (erg s$^{-1}$);
%\par\noindent 
(4) Absolute visual magnitude assuming $H_0=50,\Omega_0=1$;
%\par\noindent 
(5) Radio loudness;
%\par\noindent 
(6) Browne radio core dominance R parameter;
%\par\noindent 
(7) Energy index at 1 keV;
%\par\noindent 
(8) X-ray loudness: Two-point spectral slope between 2500\AA and 2 keV (rest frame)
}

\vvs
Table 3. X-ray spectral fits using power law model with galactic absorption.
 References:
1) This paper; 2) Wilkes and Elvis (1987); 3) Masnou et al (1992); 4) Elvis et al (1991);
5) Comastri et al (1992); 6) Turner and Pounds (1989); 7) Williams et al (1992);
8) Ohashi et al (1992); 9) Della Ceca et al (1990); 10) Turner and Pounds (1988);
11) Tananbaum et al (1986); 12) Saxton et al (1993).

\vvs
Table 4. (a) BVRI CCD photometry on the Johnson
system for \sam~quasars. (b) BVRI CCD photometry, for 
other IPC quasars. (c) UBVRI photometry from Mt. Lemmon.

\vvs
Table 5. Dereddening values used in Neugebauer
et al (1987). These values were used to rederive
the as-observed fluxes for the PG objects
so that we could apply our own dereddening correction.

\vvs
Table 6. Near infrared photometry. Magnitudes
are on the instrumental system; Vega defined
to have magnitude zero in all bands. (a) \sam~quasars
at MMT. (b) \other~quasars at MMT. (c) Quasars
observed at Mt. Lemmon.
Photometry is in 12 arcsecond beam except for those
dates marked with an asterisk which are in an 8 arcsecond
beam; see Cutri et al (1985) for observational details.

\vvs 
Table 7. Mid infrared photometry. (a) \sam~quasars;
(b) Other IPC quasars. Magnitudes as Table 9.

\vvs
Table 8. IRAS fluxes and upper limits.
AO = Additional Observation; LA = Lineadd from
survey. The value for 0915+165 is from the Point
Source Catalog and that for 1351+695 is from 
Low et al (1988).

\vvs
Table 9. Radio observations from the
literature. (a) Core fluxes. (b). Fluxes
of the extended radio source (core excluded).
%\newpage
References:
(1) Kellerman et al (1989);
(2) I. Gioia, private communication;
(3) Gower and Hutchings (1984);
(4) Unger et al (1987);
(5) Preston et al (1985);
(6) Feigelson, Isobe and Kembhavi (1984);
(7) Rudnick, Sitko, and Stein (1984);
(8) Miley and Hartsuijker (1978);
(9) Price and Milne (1965);
(10) Pooley and Henbest (1974);
(11) Perley (1982);
(12) Spencer et al (1989);
(13) Swarup, Sinha, and Hilldrup (1984);
(14) Wills (1979);
(15) Owen, Porcas and Neff (1978);
(16) Hintzen, Ulvestad and Owen (1983);
(17) Antonucci and Barvainis (1988);
(18) Hutchings and Gower (1985);
(19) Shimmins and Bolton (1981);
(20) Bolton and Butler (1975);
(21) Wills (1975);
(22) Ekers (1969);
(23) Shimmins and Bolton (1972a);
(24) Shimmins and Bolton (1972b);
(25) Meurs and Wilson (1981);
(26) Edelson (1987);
(27) Gregory and Condon (1991);
(28) White and Becker (1992);
(29) Wright et al (1991);
(30) Jagers et al (1982)

\vvs
Table 10. Millimetre wave fluxes and
upper limits.
References: 1) Robson, et al (1985); 2) Owen, etal (1978);
3) Owen and Puschell (1982);  4) Landau, Epstein and Rather (1980);
5) Ennis, Neguebauer and Werner (1982);  6) Clegg, et al (1983);
7) Chini, Kreysa and Biermann (1989); 8) Edelson, Malkan and Rieke 1987;
9) Steppe, et al. 1988; 10) Engargiola et al, 1988; 11) Antonucci,
Barvainis and Alloin 1990.

\vvs
Table 11. Magnitude scale zero points,
assuming zero magnitude for Vega. Values
are given for an object with the same spectrum
as Vega and for an object with a flat 
energy distribution (power law
spectrum $F_\nu\sim\nu^{-\alpha}$ of slope unity).

\vvs 
Table 12. Adopted luminosities and half light radii
for host galaxies, derived as discussed in the text. 
No spatial information was available for objects marked with reference (0),
and they have been assigned the median luminosity of the remainder
of the sample.
References for observational data:
(1) Boroson, Oke and Green (1982);
(2) Gehren et al (1984);
(3) Smith et al (1986);
(4) Boroson and Oke (1984);
%(5) Boroson, Persson and Oke (1985);
(5) McAlary et al (1983);
(6) Neugebauer et al (1985);
(7) Yee and Green (1987);
(8) Hutchings, Crampton and Campbell (1984);
(9) Griersmith and Visvanathan (1979).
(10) Hutchings, Johnson and Pike (1988);
(11) Hutchings, Janson and Neff (1989)
(12) Neizvestnii (1986);
(13) MacKenty (1990).

\vvs 
Table 13. Estimates of variability at 4 wavelengths.
The quantity given is the observed range of variability
(maximum value over minimum) followed by the corresponding
observational interval in years. 

\vvs 
Table 14. Bolometric and UVOIR luminosities, in units
of $10^{44}$  \mbox{erg s$^{-1}$}.

\vvs
Table 15. Decade luminosities, in units
of $10^{44}$  \mbox{erg s$^{-1}$}.

\vvs
Table 16. Octave luminosities, in units
of $10^{44}$ \mbox{erg s$^{-1}$}.

\vvs
Table 17. Bolometric correction factors
for UV, visible and infrared monochromatic luminosities.
Monochromatic luminosities are defined to be the value of
$\nu L(\nu)$ in the rest frame. Mean and standard deviation
are given, followed by the minimum and maximum values
found in the sample. Errors in the determination of
individual energy distributions have been ignored for
the purposes of this table. Also listed are estimates
of the ionizing flux discussed in the text.

\vvs
Table A1. Continuum spectrophotometry has been binned into
the rest frame bandpasses defined by this table, chosen
to avoid strong emission lines as described in the text.
Tables A1 to B47 are published on CD-ROM only in ASCII
format. Latex formatted printed versions are available
from the authors on request.

\vvs
Table A2. IUE continuum fluxes. Fluxes are
measured in bands chosen by rest wavelength
to avoid emission lines, but are given 
in the observed frame. 
%(a) IUE observations
%of the \sam; (b) IUE observations
%of other quasars.


\vvs
Table A3. Optical continuum spectrophotometry.
Fluxes are
measured in bands chosen by rest wavelength
to avoid emission lines, but are given 
in the observed frame. 



\vvs
Table A4. Extinction law as a function of
frequency, based on
Savage and Mathis (1979) and Rieke and
Lebofksy (1985). Values are relative to the
extinction at V.

\vvs
Table A5. Energy distribution of starlight template used for host
galaxy subtraction, normalized at V.

\vvs
Table B1-B47. Observed frame energy distributions for
each individual quasar.
\end{document}