From: Ariel Goobar (ariel@physto.se)
Date: Tue Apr 15 2003 - 01:15:33 PDT
On Mon, 14 Apr 2003, Greg Aldering wrote:
Hi Greg,
I am on my way to the airport thus this response is very quick
and "spontaneous". In particular, I have not (yet) had the
time to look at Rob's results. However, it seems somewhat
surprising that you argue that 1) the U-B intrinsic spread
isn't critical 2) We pick a "narrow" value that would imply that
our sample does not include what you call "outliers" in the Jha
sample.
Unless, we have evidence that we don't have "ouliers" we should use
the larger number, especially as it doesn't affect the results (as
you say)!
Furthermore, if you say that the results are insensitive to the choice, it
seems that that discussion below about "best" chi2/dof is based
on very small differences, i.e not very significant.
Ariel
> First, I encourage everyone to look at Rob's recent results showing
> that the choice of U-B instrinsic color dispersion has little
> impact on either OM,OL or OM,w. This is very nice.
>
> Since we need to quote a value in the paper nonetheless, I examined the
> residuals of the z > 0.638 SNe in our sample. In this sample are 5 P99
> SNe and 5 new HST SNe. I examined the chi-squared/DOF for both
> populations for intrinsic U-B color dispersions of 0.0, 0.043, and
> 0.095 mag. I found that 0.0 mag was too small, 0.095 mag was too
> large, but 0.043 was juuuust right. 0.043 mag gives chi-squared/DOF
> very near 1 for both datasets. Conversely, 0.0 gives too large
> a value for well-measure HST SNe, suggesting that a little color
> uncertainty does exist. These results show no redshift dependence
> of the range 0.638 < z < 0.863 covered by our data. (Note that
> I have not calculated the significance of the chi-squared/DOF for
> the less optimal values, since as I argue below, what I've done
> is good enough for this paper.)
>
> One word of warning, however, is that the residual distribution for
> these SNe is strongly bimodal. Half the SNe have very low residuals,
> peaking near 0.3, and half have residuals peaking near 1.5. The P99 and
> new HST SNe are evenly spread between the two peaks, with both samples
> having 2 SNe in the lower peak and 3 SNe in the larger peak. There is
> no redshift dependence. This a little weird given that the error bars
> are so totally different for these two datasets.
>
> In conclusion, an intrinsic color dispersion in U-B of 0.043 mag
> best describes our own observations. (One caveat is that the residual
> distribution distribution is not Gaussian for any of the above
> choices of intrinsic color dispersion.) An intrinsic color dispersion
> of 0.043 is consistent with Jha's thesis over the stretch range of
> our SNe, after two outliers are removed from Jha's sample. Rob
> has shown that the choice of intrinsic color dispersion for U-B has
> little impact on OM,OL or OM,w. Therefore, I suggest that we use
> 0.04 (or 0.043 if you like), and close the book on this chapter.
>
> Cheers,
>
> Greg
>
>
-- ___________________________________________________________________ Ariel Goobar (www.physto.se/~ariel) Department of Physics, Stockholm University AlbaNova University Center, SE-106 91 Stockholm, SWEDEN tel: +46 8 55378659 fax: +46 8 55378601
This archive was generated by hypermail 2.1.4 : Tue Apr 15 2003 - 01:16:07 PDT