From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Wed Feb 26 2003 - 15:14:48 PST
Since my comments about the intrinsic color dispersion in U-B and B-V
has generated the most feedback, let me try to clarify why I am uneasy
about the current approach. First, let me make clear that I am arguing
less about outright errors in any data, but in its interpretation in
this *very specific* context. Mine is not an airtight case, and I
believe there is some intrinsic color dispersion.
I'll try to break things into points:
I think Alex Kim came closest to capturing part of my argument when he noted
that the Lentz models say the intrinsic color dispersion due to metallicity
will increase towards the blue. For only one high-z HST SN does the F675W flux come mostly from U-band, Yet, we are taking the dispersion hit for the four
highest z SNe.
As Peter, Gene, and others have shown, a flux-limit search is heavily weighted
towards UV-bright SNe at high redshifts (quite unlike the case for Jha's
sample). Therefore, selection effects in the high-z data can severely
constrain how red an SN would be detected in U-B. This then constrains which
of Jha's/Riess/Hamuy SNe to include in getting the intrinsic dispersion.
You can see this in action by looking at our stretch distribution, for
U-band-select SNe which all have high stretch. If you go to Jha's U-B
vs stretch plot, you see these are the bluest SNe. And, you might agree that
if you place a prior on Jha's stretch and color appropriate for a
high-redshift search simulation, the dispersion is probably smaller (for
instance, you strongly deselect againt low stretch and/or red SNe.
I am using this argument at high redshift only because without the selection
effects, perhaps you see more SNe with U-B like Jha's. However, I would argue
that the fact that Jha's reconstructed bandpasses aren't great matches to
to real UBVRI likely means that there are still problems with the phomotery
at the level of a few percent (which we will face for Spring '99). Thus, I am
not comfortable assuming that the remainder after taking out the
quoted statistic errors applied at high-redshift. The point may be moot at
low redshift, but not at high redshift where HST was used.
A relative item is that in the Nobili et al paper, it appears that Hamuy and
Riess give very different intrinsic uncertainties. I think this supports my
concerns above.
Finally, I think that the too-good chi^2 for the highest z HST SNe supports
my overall argument.
- Greg
This archive was generated by hypermail 2.1.4 : Wed Feb 26 2003 - 15:14:48 PST