Re: Fixed vs. floating offsets

From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Wed Feb 19 2003 - 14:25:15 PST

  • Next message: Lifan Wang: "Re: Fixed vs. floating offsets"

    On Wed, Feb 19, 2003 at 01:59:59PM -0800, Alex Conley wrote:
    > I've refit a couple of Rob's troublesome low-z sne with and without

    OK, here's one example 1992ag -- my fits, my K-corrections. I've
    attached two plots, one is with floting zero, the other is without
    floating zero. (Look at the plot to figure out which is which.) You
    can't tell me that the fixed-zero fits are reasonable. Using that in
    cosmology would be inexcusable. Notice that m_B goes from 17.38 to
    16.57 when I go from a fixed to a floating zero. That's huge. Looking
    at the data and the fits, it's *very* clear which of the two of these I
    ought to believe-- the one with the floating zero.

    (The zero offset is only 0.03 compared to the flux height of the
    lightcurve, so it's not a big offset being fit here; it's just all those
    low-down points distorting the lightcurve. Why didn't Alex see this
    huge difference? He eliminated points later than 30 days after max.
    However, for purposes of this paper, it would not be a good idea to just
    eliminate those later points, because on some of our high-z supernovae,
    we have points, including HST points, that late; for instance, consider
    SN9855, whose lightcurve plot I've also attached-- and that's not even
    the one with the latest HST points.)

    (Part of the problem also is that I suspect Hamuy has lied about his
    error bars, by quoting them as magnitude error bars, giving the low
    points more signifiance than they observe. But, we can do no better
    than to trust the error bars published.)

    I have similar trouble in the other direction if I try to *float* every
    zero-- plus the issue that I discussed in the meeting, that I've already
    implicitly assumed that the zero has been proprely subtracted when
    combining HST and ground data.

    There is no "right" way to do this that is utterly pure as the driven
    snow statistically. Treating every supernova in a purely consistent
    manner gives results that are obviously plain wrong, as the attached
    supernova demonstrates. So I picked the fit procedure that worked best
    for most of them, and then changed it for those few supernovae where it
    was necessary to do so. This is the best we can do.

    -Rob

    -- 
    --Prof. Robert Knop
      Department of Physics & Astronomy, Vanderbilt University
      robert.a.knop@vanderbilt.edu
    






    This archive was generated by hypermail 2.1.4 : Wed Feb 19 2003 - 14:25:40 PST