late-time fitting debate

From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Sat Mar 15 2003 - 12:37:38 PST

  • Next message: Robert A. Knop Jr.: "Re: 97aj"

    Hi all,

    Before the discussion of the late-time fitting gets out of hand, let
    me try to introduce some new information into the debate.

      1) First some history:

         In P99 we cut points after +60 restframe days post-max. According
         to my inspection of the fitting files and the text of P99, we did
         not deweight!

         In Goldhaber 2001 we cut points after 40 restframe days post-max
         and seem to have put in an error floor of 0.007 (The only place
         I could find where this error floor is mentioned is in the footnote
         to Table 7 of that paper!).

         These are published procedures (although the deweighting is not
         clearly presented) that we can adopt and cite without further
         justification, if desired.

      2) Since P99, in particular in AKN00, we generally agreed that stretch
         begins to fail and our knowledge of SN lightcurve variety is poor
         starting 25-40 days after max. This argues for the concept of an
         uncertainty on the TEMPLATE (as Saul has suggested).

      3) We know that if we cut at (1+z)*40 days, following G01, we will lose
         some HST lightcurve points. However, if we cut at (1+z)*60 days,
         following P99, we should be OK.

      4) In the case of the nearby SNe, we know that final reference
         images were not always used to obtain the galaxy contamination
         level. We do not know for which SNe references were or were not
         used, and we don't know the uncertainty to associate with galaxy
         missubtraction. For that reason, there is additional justification for
         cutting or deweighting the nearby late-time points despite the quoted
         uncertainties.

      5) Note that at 50 days the B-band template is a f/fmax = 0.054, so
         even this late the error floor of 0.007 would affect points with
         claimed S/N > 8.

    Putting this altogether, I suggest that we cut points after 60
    rest-frame days (but not final refs of course) as in P99, and add in an
    error floor, with the error floor to be interpreted as accounting BOTH
    for uncertainty in the template shape and stretchability at late epochs
    and for some uncertainty in the final-reference subtraction for nearby
    SNe. We can cite this a combining the treatments presented in P99 and G01.

    - Greg



    This archive was generated by hypermail 2.1.4 : Sat Mar 15 2003 - 12:37:58 PST