Re: comments on Oct 12 manuscript

From: Alexander Conley (AJConley@lbl.gov)
Date: Mon Nov 08 2004 - 13:11:51 PST

  • Next message: Alexander Conley: "Re: comments on Oct 12 manuscript"

    Hi Greg,

       I've been looking at the color distribution at maximum. I guess the
    first reaction
    should be: what in the world was I thinking, saying that those two
    distributions
    were compatible -- they don't even really look like it to the eye.

    And, in fact, I did discover than using the very weak color cut (0.6)
    for
    the max mag fits was disastrous -- the Hubble diagram looked terrible,
    and
    the results were way off of expectation. I ended up using a cut of 0.2,
    and things seemed to work well.

    So I tried to look at the effect of various color cuts on the mean
    colors.
    A complicating factor is that there are two high-z SNe with rather
    anomalous
    colors: 01iy, and 97af (-0.32, -0.24). Now, 97af has huge errors, so
    may
    not be a big deal, but 01iy claims to have good errors. Looking at the
    actual lightcurve, it is rather poorly sampled. I'm not sure it's
    wrong, but
    I think I can categorically state that this is one of those cases where
    snmin
    radically underestimates the errors. Incidentally, removing either of
    these
    SNe from the sample does not particularly affect the cosmological
    results
    (see the jackknife tests).

    Okay -- so here's how it plays out in mean color:

      E(B-V) < 0.6 : low-z: 0.09417 +- 0.0248 high-z: 0.0229
    +- 0.0287
         +01iy removed: low-z: 0.09417 +- 0.0248 high-z: 0.0378 +-
    0.0256
         +01iy,97af removed: low-z: 0.09417 +- 0.0248 high-z: 0.0504 +-
    0.0234

    E(B-V) < 0.4 : low-z: 0.07 +- 0.0191 high-z:
    0.0229 +- 0.0287
        +01iy removed: low-z: 0.07 +- 0.0191 high-z: 0.0378
    +- 0.0257
        +01iy, 97af removed: low-z: 0.07 +- 0.0191 high-z: 0.0504 +-
    0.0234

    E(B-V) < 0.25 : low-z: 0.04516 +- 0.0140 high-z:-0.0009
    +- 0.0257
       +01iy removed: low-z: 0.04516 +- 0.0140 high-z: 0.0143 +-
    0.0217
       +01iy, 97af removed: low-z: 0.04516 +- 0.0140 high-z: 0.0270 +-
    0.0185

    E(B-V) < 0.2 : low-z: 0.03933 +- 0.0132 high-z:-0.0124
    +- 0.0241
       +01iy removed: low-z: 0.03993 +- 0.0132 high-z: 0.0030 +-
    0.0195
       +01iy,97af removed: low-z: 0.03993 +- 0.0132 high-z: 0.0158 +-
    0.0152

    E(B-V) < 0.1 : low-z: 0.04091 +- 0.0097 high-z:-0.0344
    +- 0.0243
       +01iy removed: low-z: 0.04091 +- 0.0097 high-z:-0.0176 +-
    0.0187
       +01iy, 97af removed: low-z: 0.04091 +- 0.0097 high-z:-0.0038 +-
    0.0133

    Whew!

    Okay -- ignoring 97af, 01iy, what does this say?

       Well, with E(B-V) < 0.6, the low-z sample is about 0.043 mag redder,
    which
      translates into 0.18 mag for Bmax and 0.09 for CMAGIC. This reddening
      is detected at 1.28 sigma.

       With a slightly tighter cut, E(B-V) < 0.4, the low-z sample is 0.02
    redder,
      or 0.08 for Bmax and 0.04 for CMAGIC. This reddening is detected at
      0.64 sigma.

       Tighter still, E(B-V) < 0.25, the low-z sample is also about 0.02
    redder, with
    the same consequences. This is at 0.78 sigma.

       Tightening to E(B-V) < 0.2, the value used for the maxmag fits, the
    low-z sample
    is also 0.02 redder, at 1.18 sigma. Again, this is 0.08 in Bmax and
    0.04 in CMAGIC.

       Finally, going to E(B-V) < 0.1, the low-z sample is redder by 0.008
    at 0.47 sigma.
    This is negligible.

    There is a difference, but it seems to be heavily driven by the two
    really blue points,
    01iy and 97af.

    One may view this as an argument that a tighter extinction correction
    should be used.
    I have already done this experiment in the method paper. Here are the
    results,
    quoted in the principle axis frame of the error ellipse. Recall that
    the error along
    the short axis is about 0.11, and along the long axis is about 0.82.

    Given in terms of shifts relative to the primary fit
    E(B-V) < 0.4: -0.007 short, +0.033 long -> negligible
    E(B-V) < 0.25 : -0.038 short, +0.0426 long -> about 1/3 sigma short,
    negligible long
    E(B-V) < 0.1 : -0.0018 short, -0.547 long -> about 1/2 sigma long

    The interesting one is the really tight cut, which moves the error
    ellipse strongly
    along the long axis. This is one of the major long-axis systematics
    left in this study.

    Bottom line:
      The low-z sample is redder than the high-z sample. This messes up the
    maximum
    magnitude fit, but doesn't do nearly as much to the CMAGIC fit.

    Alex



    This archive was generated by hypermail 2.1.4 : Mon Nov 08 2004 - 13:13:04 PST