From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Mon Feb 24 2003 - 10:46:19 PST
Preface: First let me say that I agree that the newest draft of the
HST paper is a dramatic improvement and it seems that we should be
able to finalize it quickly. As I haven't been able to read comments
from everyone else, I apoligize for redundant comments. Further
caviats are that I am typing on a French keyboard so generally
translate "q" to "a", and I am running a fever so may be a little
more lazy that usual in going into detail.
Overall:
The theme of this paper is HST makes better measurements which
allow an independent cosmology measurement from 4x fewer SNe. However,
given the assumed intrinsic color uncertainties there are 5 P99 SNe
with E(B-V) which is as good as the 4 highest redshift SNe in this
paper since the intrinsic uncertainty appears to domonate. Thus, one
could conclude that one should not have pursued SNe with z > 0.6, from
space or ground. The current paper could cure this problem by saying that
many of these SNe were observed with NICMOS, and that a future analysis
will present the complementary and improved colors based on B-V and V-R.
Photometry:
I think something more than the citation to Dolphin is needed to
reminder people of the main issues with HST photometry.
Readers like me would like to know the status of the calibration,
and what aperture corrections were used/calculated/assumed. (The
HST Key Project - H0 papers are good sources for this info and can
be cited.)
Demonstration that the HST and ground are on the same system, e.g.,
using fiducials, is important. The fits suggest that the new HST data
are giving fainter SNe than P99, and the question is whether we know
that this is not due to photometry offsets. You can rest assured that
lack of such a demonstration will be cited as a weakness of this paper,
yet it seems like something that is not hard to check. I hqve seen that
error terms have been included to account for such changes, but there
is no explainqtion for the basis of those errors (so I assume they
are only guesses).
Mention that O'Donnell extinction law is an *improvement* over CCM,
and give some idea of the size of the differences. Also, since SDF
probably used CCM to convert from FIR surface brightness to dust column,
it is notclear to me that you can now switch from CCM to O'Donnell
for Galactic extinction.
I couldn't tell whether HST final references were used to constrain the
a_j's; please clarify.
Do you say that you marginalized over the photometry nuisqnce parameters,
like x0,y0. We should comment that we did enforce the position to be fixed
with respect to the image coordinates due to the lack of many reference
sources and the complex and time-dependent nature of the PC coordinate
transformations (cite Anderson & King in a recent issue of PASP).
Then there is the question of whether we should show lightcurves - I think
for this paper we should. Since the referee will likely ask for them,
we could delay that work until after submission (although we should all
get a look at the lightcurves and fits internally).
Comparison with old analysis:
For internal discussion I would like to see the H0-free distance modulii
from the current analysis plotted versus that from P99. I'd like to see
nez versus old color plots too.
It is claimed that the color excesses in P99 used the Nugent template,
however for B-V this was not the case.
We should point out that in P99 we used Bmax-Vmax, not (B-V)max as in
the HST paper
Consistency in analysis:
From what I have gathered, Rob is using late-time points for nearby SNe
because some HST lightcurves go to late times. I understand this rationale,
but if there are peak/tail differences (e.g., as in AKN00) then this
rationale falls apart because the nearby lightcurves fits will be dominated
by this systematic whereas late-time HST points will still be dominated
by statistical errors. Fitting a baseline can appear to fix this problem,
but unless the ligthcurve at max is already well-constrained its value will
be in error by roughly the amount of the offset. If we have to use late-time
lightcurve points, then there should be some kind of correlated error in
the lightcurve. I think that a reasonable away to do this is a add a constant
correlated error normalized by the flux at peak for t > 25 days. The size of
this error can be set to get chi-squared ~1 at late times, or more crudely,
using the scatter in the fitted baselines.
Intrinsic color uncertainty:
This really hits us hard, especially in U-B. I am not yet convinced that
intrinsic color dispersion has been proven, at least not for B-V. For instance,
can it be demonstrated that variations in the dust extinction law are not to blame?
This is a big difference for the errors, since an intrinsic color error is
multiplied by R_B while an error in R_B is multiplied by E(B-V), the latter being
much smaller. I looked at the Nobili et al paper, and also Jha's thesis, and have
my reservations about putting this error in at its current size. Also, since we now
have the Jha data we should at least use U-B based on his data since it overwhelms
the other available data. (Here again, since the Jha data are now available, this
doesn't sound to me like it has to cause much of a delay.)
Misc:
Please use real IAUC names for our high-z SNe
Note that R99 is expected to have more significant extinction since it isn't
a flux-limited sample and can thus detect more extincted SNe but doesn't add
less-extincted SNe at higher redshift. You can cite Gene's draft paper or my
discussion in Cosmic Explosions. This is why the R99 dataset was not appropriate
for the Fit C / no extinction correction analysis of P99 (and really isn't
appropriate for Fits 1-3).
9784 / SN1997ap has no host to quite faint limits, so I think one
can argue that it doesn't require extinction correction.
Can you clarify whether alpha was allowed to take different values with and
without extinction-correction to the SNe? This seemed to be a never-ending
source of confusion from P99 and it would be good to fix that here if possible.
I suggest using a 2-sigma cut on the detection of reddening for thefits
which are uncorrected for extinction. We aren't trying to prove that extinction
exists, and at 2-sigma the chance of a false rejection is still very low while the
bias due to extinction is decreased by 1-sigma (i.e. several hundredths of a
magnitude).
Minor global fixes:
chi-squared looks better as $\chi^2$
1-sigma looks better as $1\,\sigma$ (likewise for 2 and 3 sigma)
I like $U$ rather than U, for filter designations. This helps with $I$
especially. For colors I like $U$-$B$ even though $U-B$ is technically
correct.
Specifics:
page 4; para "Each of ..."
This needs rewording. The filters were chosen to maximum sensitivity to these
faint objects (precluding use of the F850LP filter) and to be as close as practical
to restframe B and V at the targeted redshifts. These filters approximate
Kron-Cousins R- and I-band. ALso gie a reference to the HST WFPC2 handbook
or something like that.
page 4; para "The HST ..."
When discussing traditional PSF photometry be sure to say that it is optimized
to photometrer groups of points sources. (This is really why it was invented -
for isolqted point sources you can just apply a radial weight.)
Also, 7x7 and 9x9 don't say that these are pixel regions
page 6; para "Fluxes ..."
Last sentance sounds like you are trying to justify an unconventional approach -
can you reword this so it sounds like you are doing something quite normal
(or if I have missed the point, please state why this assumption is OK)?
Section 4
WMAP results need to be included. Also, when you mention the motivation for
a flat universe, dont't forget about inflation!
Lots of very little things that others may have caught, or which I can fix
in the next version.
Overall, this looks like a week at most for a very polished paper. I think we
also need a little more discussion/documentation internally so we all really
understand the analysis and can stand by it - hopefully less than a week for
that.
Cheers,
Greg
This archive was generated by hypermail 2.1.4 : Mon Feb 24 2003 - 10:46:20 PST