From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Sun Mar 23 2003 - 14:45:01 PST
See:
http://brahms.phy.vanderbilt.edu/~rknop/scp/hst/#screwingaround
Based on a suggestion from and discussion with Greg, I tried doing the
fit where instead of fixing alpha for purposes of error propogation, I
propogated the stretch error into magnitude errors using the alpha
currently being tested. This means bigger error bars at bigger values
of alpha, which will tend to drive the fit to want bigger values of
alpha since it gets lower chisquares due to the larger error bars.
Line B-4 shows the results on the P99 data, using both the data dirctly
from that gersontable and my refits/re-Kcorrections of it.
Interestingly, with a fixed alpha for uncertainty purposes, my
lightcurve parameters tended to return a higher best-fit alpha than did
the P99 lightcurve parameters. We get the opposite result when the
alpha used for error analysis is the same as the alpha being tested for
each chisquare value. I don't really know what to make of that
observation, though.
There are all sorts of worries associated with systematic effects and so
forth (specifically, you're more likely to find high-stretch supernovae
in a little more host galaxy extinction, so if you aren't correcting for
extinction you'd expect that to lower alpha... but it gets really scary
when you think about comparing that systematic in the low and high
redshift subsets).
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Sun Mar 23 2003 - 14:45:03 PST