Re: Evolution paper

From: Gabriele Garavini (garavini@in2p3.fr)
Date: Tue Jan 25 2005 - 12:58:34 PST

  • Next message: Gabriele Garavini: "Evolution paper New Draft."

    > What Eric is currently seeing with code that we *think* reproduces what
    > you've done is that the high-z supernovae he has from Keck are showing
    > up with EW's a little bit lower than the ones you have-- sort of at the
    > ridgeline of your distribution. It is possible that this results from
    > systematics in different galaxy subtraction. Eric is using Andy's fit
    > program for galaxy subtraction; what are you using?

    This is in principle possible. We allowed for this source of systematic
    and extra uncertainties of 10% in the galaxy subtraction. This is
    included in the paper and added to the statistical error. The procedure
    used to subtract the galaxy is that written by Gregory Sainton, and
    uses a
    similar approach as that of Andy, but with a different minimization
    algorithm.

    >
    > Can you send us the galaxy-subtracted spectra from which you made your
    > high-z measurements? (The original galaxy unsubtracted ones are all
    > the
    > ones on Isobel's spectrum page, yes? If not, please send us the
    > location where we can find those.) That way, Eric can try his galaxy
    > subtraction and compare it to your galaxy subtraction on the *same*
    > spectra to see if that is the systematic. It may also allow us to
    > diagnose if Eric's procedure really is the same as your procedure
    > without having to wait for Gaston's numbers (although we *still* *need*
    > *those*, so please get them to us!)

    please find the tar file with the vlt_hz spectra at
    supernovae.in2p3.fr/~garavini/papers/vlt_hz.tar

    >
    >> Rob do you find the new section of the Evolution paper where I explain
    >> the technique and possible systematic, clear enough?
    >
    > A couple of outstanding issues:
    >
    > * What procedure do you do to identify the maximum, i.e. the center
    > of
    > the 20-30 angstrom window on each side? In low-z spectra, it's
    > visually obvious, but it's less so with high-z spectra. You should
    > comment how you choose it. Is it just by eye? Are you smoothing
    > and doing it by eye? Are you fitting a parabola to the local data?
    > Or something else? (Eric speaking: this is currently my biggest
    > question about the procedure.)

    Unfortunately I'm still waiting for the exact number of the window size
    in
    the high and low S/N ratio cases, but what is written in the paper right
    now should be reasonable. Still I'd like to have the final number from
    Gaston. For the procedure: I think Gaston was actually looking at the
    unsmoothed spectra by eye and was not fitting any parabola.

    >
    > * Can we confirm the 20-30 angstrom region? Why not just pick 25
    > angstroms and always use that for noisy spectra? (This is what
    > Eric
    > has done so far.) If sometimes you use one window and sometimes
    > another window, what is it that guides the choice?

    As before, this is a detail Gaston need to sort out.

    > * It's still a little unclear exactly what you're doing. One
    > paragraph
    > says straight line fit. The next paragraph says the wavelength
    > regions selected. I would suggest the following wording for
    > Section
    > 3.2.1, starting with paragraph 2:

    > In order to measure an equivalent width (EW), the wavelength
    > range of the line and the underlying continuum must be
    > identified. In the case of narrow emission or absorption
    > lines,
    > this procedure is straightforward; however, SN spectra show
    > very
    > broad features bounded not by smooth continua, but mearly by
    > local maxima between them. The wavelength range and continua
    > used for EW measurements are defiend as follows.

    > First, on either side of the feature a local maximum is
    > identified; these local maxima are used as the edges of the
    > wavelength range in which the equivalent width is measured.
    > Table 4 lists the range of wavelengths in which the local
    > maxima
    > on either side of each feature can be found. In order to
    > specify the location of the local maximum, we DO WHATEVER IT IS
    > THAT WE DO TO FIGURE THIS OUT.

    > Once the positions of the local maxima are identified, we
    > extract all of the data within a small window around both
    > maxima. To those data, we perform a chi-square fit of a
    > straight line. This fit is used to define the effective
    > continuum for the equivalent width measurement. (This is an
    > empirical definition used for purposes of the measurement;
    > given
    > the complicated blended nature of the SN spectra, it almost
    > certainly does not represent a "real" continuum.) The
    > wavelength span of these windows depends on the SNR of the
    > spectrum. In high SNR spectra, a typical window has a width of
    > 10 A (?????). In low SNR spectra, where the position of the
    > maximum is less easily identified, a larger region (typically
    > 20-30 angstroms, depending on WHAT DOES IT DEPEND ON?) is used.
    >
    > The systematic effect of low signal to noise... (((last two
    > paragraphs as is for now))).
    >
    >
    > * Please explicitly say that you're using the original resolution of
    > the spectrum without smoothing or binning, if that is the case.
    >
    > -Rob & Eric
    >

    I think the rewording sounds good to me. I'll put it in as soon as
    possible.

    Thanks
    Gabriele

    -- 
    ======================================================================
    LPNHE - IN2P3 - CNRS University of Paris VI and Paris VII
    4 place Jussieu, Tour 33 - Rez de chaussee 75252 Paris Cedex 05 France
    Phone: +33 1 44 27 41 54,  e-mail: garavini@in2p3.fr
    ICQ: 148161845, AIM: gabrigaravini
    


    This archive was generated by hypermail 2.1.4 : Tue Jan 25 2005 - 13:53:19 PST