From: Lifan Wang (lifan@panisse.lbl.gov)
Date: Mon Mar 17 2003 - 10:04:43 PST
> From owner-deepnews@listserv.lbl.gov Mon Mar 17 09:28:13 2003
> X-Authentication-Warning: listserv.lbl.gov: majordom set sender to owner-deepnews@listserv.lbl.gov using -f
Don,
We should really discuss this off-line. This topic is only indirectly
related to the current HST paper and concerns mostly of future l.c. fits.
So anyone who is not interested in these details can skip this message.
Lifan
> To: "Robert A. Knop Jr." <robert.a.knop@vanderbilt.edu>
> cc: hstpaper@panisse.lbl.gov, <deepnews@lbl.gov>
> Subject: Re: late-time fitting debate
> MIME-Version: 1.0
> Content-Type: TEXT/PLAIN; charset=US-ASCII
> Sender: owner-deepnews@lbl.gov
>
>
> > On Sat, Mar 15, 2003 at 01:03:04PM -0800, Lifan Wang wrote:
> > > So my preferred approach is to do stretch fit only
> > > for data before nebular phase.
>
> Lifan, what do you mean? What sort of s(t) do you imagine? I'm left with
> the stupid picture of a lightcurve with a discontinuity at the transition
> to the nebular phase. I could even think about implementing it, but, as
> Rob says, not for this paper.
>
This is exactly what the S-Stretch that I have developed. I am not advocating
this for this paper, but what I learned from S-Stretch fit is that Hamuy
OVERestimated his errors bars rather than underestimated.
> In addition to the 56Co problem, there's the question of whether late
> supernovae follow the same template, the question of how good our template
> is out thar, and questions about galaxy/background subtractions by the
> (non-SCP) people that reduced the data. My impression in looking at the
> Hamuy data a long time ago is that late times have all sorts of problems,
> maybe physics and maybe data reduction.
>
It is not because the errors of the template. For most SNe, late time is given
by the extremely simple exponential law, the template can in fact be very good.
Our problem is all about stretch. At day 60, a stretch of 1.1 practically
increases the half-life time of 56Co by 6 days, and the later the epoch
the more serious the problem - this is simply not acceptible for a good
fit. Indeed Hamuy's l.c. has a lot of problems at late phase. For some SNe,
the problem is obvious as it apparently deviates from an exponential law.
The 0.03 mag assigned to all sorts of systematic errors seem to be
reasonable for the entire data sample, although some SNe such as SN1992ag
may have an error somewhat larger than that.
So there is hardly any good reason for re-inventing
the errors. If an SN such as SN1992ag fails stretch fit, it may
mean two things: (1) Stretch does not apply; (2) The SN data is wrong.
SN data mean both the central values and their errors - *modifying errors
is not very much different from modifying the central values*. But if
it can be justified, the model can be modified or deviant data points
can be rejected. Rejecting bad data points has some subtle statistcal
consequences but is acceptible in general.
The dates for which both stretch and data agree can
be found in Gerson et al (2003, in prep).
> What do you suggest?
We can talk more about this offline.
>
> Don
>
This archive was generated by hypermail 2.1.4 : Mon Mar 17 2003 - 10:05:06 PST