From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Sat Nov 06 2004 - 23:48:04 PST
Hi Alex,
Here are my comments on the main issues of the Oct 12 manuscript of the
CMAGIC paper. Here I mostly focus on issues which might affect the blind
analysis, reserving minor editorial commments for later. In the end, I
did find a few things which you will want to look into before unblinding
the analysis. The checks are easy, but one of the results might require
refitting.
* As you have noted in follow-up to my original query, there may be a trend
in the scatter of the Bmax residual as the CMAG residual increases. As
there is no clear reason for this, and the affect is weak, you may simply
wish to note this in the text or caption. (Have you thought about
whether bumps could play a role?)
* You have confirmed my impression the the RMS scatter about the Hubble
line is smaller at high redshift than at low redshift (by about sqrt(2)).
This should be discussed. Again, I don't think it would affect the blind
analysis since there is a plausible explaination for this effect. Any
evidence that extinction+selection effects could be the cause should be
presented to the reader.
* I had asked whether a Ib/c would follow the CMAGIC relation. I did not see
your response to this question. If Beta for a Ib/c is different than for
a Ia, it might present an objective and self-consistent test for non-Ia
contamination in your sample. This could affect the sample membership, and
thus the unblinded results.
* Motivated by possible offsets seen between Hamuy, Riess and Jha in some
of Serena's early work, I had asked you to provide the mean residual (and error
in the mean) for each subset. I also requested this be done for other sub-samples,
such as SNe which are fit as V-->B, R-->B, or I-->B. I didn't see these mean residuals
quoted. This might be an obvious place to mention the Knop/Barris offset.
As the mean residuals for the above subsets should be easy to calculate, I
would like to see the values and their uncertainties before the blinding is
taken off.
* In the caption to Figure 4 I had asked for a test of the relative (B-V)max
distributions for the low- and high-redshift samples. You quote the sample
means, but it appears that you quote the RMS about the mean rather than the
error in the mean.
If my guess is correct, then the low redshift sample has a mean of
0.045 +/- 0.028 and the high-redshift sample has a mean of -0.037 +/- 0.022.
The difference of 0.082 is significant at 2.3 sigma.
Is the point at (B-V)max = -0.4 included? Does the difference remain if
it is excluded?
If attributed to reddening, this difference implies an error of 0.34 mag between
high and low redshift for a Bmax fit without extinction correction. This
really could affect your results. If it doesn't, I'll be quite surprised.
You may have to consider having a fit which excludes the very blue high-redshift
point as a way of demonstrating the bias or lack therefore.
The (B-V)max distribution is tighter at high redshift. This likely ties in
with the smaller Hubble-fit RMS at high-redshift, discussed above. It would be
good to tie these pieces of evidence together.
* For Figure 6 you provide a comparison of the means of the Beta distributions:
<Beta> = 1.946 +/- 0.057 for low redshift and <Beta> = 1.966 +/- 0.094
However, if these are errors in the mean I don't understand the scaling.
The low-redshift sample has 8x as many objects, so for comparable Beta
distribution functions the error in the mean for the low-redshift sample
should be smaller by a factor of 2.8; instead it is only smaller by a factor of
1.6. Alternatively if these are errors in the mean, the implied RMS
is about 0.3 in Beta for the low-redshift sample, but in the caption to
Figure 2 you quote an RMS of 0.159 for the low-redshift sample.
So, could you please clarify these values?
* I asked whether prob(bump) is correlated with Beta. I didn't find the answer.
* In your tables you quote heliocentric redshifts. Do you correct these to the
CMB frame when performing the fits? I didn't find any indication that this
was done although your analysis memo suggests it probably was done. This could
affect several aspects of the analysis, so the text should be clear (and
provide the proper citations).
* Figure 2 has 44 low-z SNe and Figure 4 has 39 low-z SNe. I didn't catch the
cause for the difference.
- Greg
This archive was generated by hypermail 2.1.4 : Sat Nov 06 2004 - 23:48:07 PST