From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Sat Mar 15 2003 - 03:42:49 PST
On Fri, Mar 14, 2003 at 09:22:30PM -0800, Lifan Wang wrote:
> Just explain why a star which does not have the unfortunate history of
> been observed as a supernova at max may have a better chance of
> getting smaller errors by increasing the exposure time whereas such
> luck is impossible for a supernova that was discovered and observed
> well at max.
Lifan, you're being obstinate. Of *course* you can get smaller errors--
smaller *flux* errors-- by increasing your exposure time for later
supernova points.
But when the supernova is 0.05 its max flux, you have to expose for 400
times as long to get the *same* magnitude error. Is this usually what
is done?
Nonetheless, if you look at a lot of published data, they quote similar
magnitude errors at peak and after max. Add to that Gerson's empirical
experience, plus the empirical experience that the late-time lightcurve
(esp. with very small error bars) can formally constrain the fit values
much more than they ought to (esp. given our true knowledge of late time
lightcurve behavior). Two things are going on: we probably don't have
the best late-time fitting procedure, *and* I suspect that some people
in the literature are under-quoting their error bars at last time.
What to do? Adding a certain number of magnitudes to the points doesn't
make sense, because error bars in magnitudes for lightcurve points don't
really make sense in the first place anyway.... Do you add 10% of the
flux to the error? Well, if you truly, honestly believe that the
exposure times were 1/(flux^2) always, then, sure, that makes sense.
What we're trying to do here is find some way of softening the errors
where we both think it most likely they're most seriously underquoted,
and also where we want it to have less of an influence on our fit. Both
of those criteria are the late time lightcurve points.
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Sat Mar 15 2003 - 03:43:35 PST