Re: Bias of our low-extinction method

From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Fri May 23 2003 - 10:27:33 PDT

  • Next message: Don Groom: "Re: Bias of our low-extinction method"

    Hi Ariel,

    Although I have not tried to reproduce the Phillips result using his
    methodology, several years ago Alex Kim and I got comparable results when
    applying extinction corrections to Bmax-Vmax based on stretch fits to
    Hamuy and other local SNe. We routinely got intrinsic dispersions about
    the Hubble line of around 0.11 magnitude. We did not apply a color cut for
    that work, but I have a feeling that the SCP Hamuy subsample does not
    include as large a fraction of red objects as in the full Calan-Tololo
    sample. The value of the intrinsic dispersion does rely on accurate
    photometry errors of course, but Hamuy's quoted photometry errors would
    have to be gross overestimates to have an effect on the intrinsic
    dispersion at the 0.1 mag level (e.g., giving 0.11 mag when the
    dispersion is really 0.21 mag as would be needed if the colors scatter by
    0.05 mag).

    Now, we have heard that Jha is getting a higher dispersion. But his sample
    is dominated by very low-redshift SNe so his error bars are dominated by
    galaxy peculiar velocities.

    - Greg

    On Fri, 23 May 2003, Ariel Goobar wrote:

    > Hi Greg,
    > Have you tried to reproduce Phillips' result? As you
    > know, he got a spread in the fitted max(B-V)~0.03 only after we selected
    > the SNe that had extremely small scatter in B-V at late times... Clearly
    > a "special" subsample of SNe. I don't think I know enough to exclude
    > the possibility I suggested, but maybe you know more.
    >
    > Cheers,
    > Ariel
    >
    > On Fri, 23 May 2003, Greg Aldering wrote:
    >
    > >
    > > Hi Ariel,
    > >
    > > You're right that the other steps subsequent to discovery will have
    > > important effects. In fact, I think that the reason why we see relatively
    > > more extincted SNe at low redshift from the 1998 search is not only due to
    > > the survey flux limit allowing them to be found, but also because their
    > > observed magnitudes were similar to those of unextincted high-redshift SNe
    > > - which were the ones we were really after - and we didn't have time (due
    > > to weather) to check out all the candidates. So, these additional affects
    > > can both help and hurt us in terms of allowing extincted SNe into the
    > > sample. Unfortunately, I have no means of recreating these in a
    > > simulation because of the important subjective component. This was also a
    > > limitation in what could be said about Malmquist bias for this dataset.
    > >
    > > As for the intrinsic color dispersion of 0.05 mag, I haven't tried that.
    > > I've read Serena's paper, but have to say I'm not really convinced of that
    > > color dispersion, since if it were present one could not get an intrinsic
    > > scatter of 0.11 mag about the Hubble diagram after extinction correction.
    > > However, I will grant that there many be bonifide color outliers, since
    > > Phillips had to cut out objects with B-V > 0.2 to get his nice dispersion.
    > > So, a Gaussian core with sigma < 0.05, with some non-Gaussian outliers -
    > > that I would believe. But then throwing those all together and calling it
    > > a Gaussian dispersion doesn't seem the thing to do. The fact that Phillips
    > > could make a color cut for extreme color and then get a small scatter
    > > suggests to me that the outlier population gives itself away, and is not
    > > hiding within a Gaussian distribution.
    > >
    > > Cheers,
    > >
    > > Greg
    > >
    > >
    > > On Fri, 23 May 2003, Ariel Goobar wrote:
    > >
    > > > Hi Greg,
    > > > that sounds interesting. Have you tried what happens when
    > > > adding a Gaussian intrinsic spred in B-V around 0.05 mag?
    > > > As you point out the effect will be highly dependent on the
    > > > the flux limit you use in your simulations. I am not sure
    > > > there is a perfect 1-to-1 correspondence with the
    > > > parameter you used (dim-cut). There are a few steps between
    > > > candidate discovery (a different stages in the I-band LC,
    > > > stretches, spectroscopy screening, K-corrs, LC fit, etc) and
    > > > the residual showing up in Rob's Hubble diagram.
    > > > Ariel
    > > >
    > > > On Thu, 22 May 2003, Greg Aldering wrote:
    > > >
    > > > >
    > > > > I have simulated the behavior of our low-extinction method. I started
    > > > > with the models of Hatano, Branch and Deaton, which provide the
    > > > > probability distribution function of B-band extinction, A_B, for SNe
    > > > > observed through spiral and elliptical galaxies viewed at various
    > > > > orientations. I generated random values of A_B following this
    > > > > distribution function, but adding a flux-limit bias which depresses the
    > > > > probability of finding extincted SNe in proportion to the volume of
    > > > > space in which they would be accessible in a flux-limited sample. I
    > > > > next "observed" them by randomly chosing an E(B-V) error bar and then
    > > > > generating a Gaussian deviate with this error bar, and adding the
    > > > > result to the generated value of A_B. I then applied the cuts a SNe
    > > > > currently must pass to be included in the low-extinction subset in the
    > > > > HST paper:
    > > > >
    > > > > sigma R-I < 0.25
    > > > > E(B-V) < 0.1 or E(B-V) < 2*(sigma E(B-V))
    > > > >
    > > > > I also added a too-faint cut, requiring that SN not be fainter than N
    > > > > times its magnitude uncertainty (including an intrinsic error of
    > > > > 0.17). In Rob's last fits, all the residuals were within 3-sigma for
    > > > > the low-extinction subset, even though no cut on deviation was
    > > > > applied. Thus, my cut considers the likelihood that if there were a
    > > > > significantly dim outlier we might have chosed to reject it even
    > > > > without a well-measured color to prove that it is reddened.
    > > > >
    > > > > I find that our method does have a bias, but that the bias is small
    > > > > compared to our other systematics. However, it might not be small
    > > > > relative to the Riess prior (that has to be checked).
    > > > >
    > > > > Here is what I get:
    > > > >
    > > > > N-sigma <A_B> <A_B> Bias
    > > > > dim cut low-z high-z low-high
    > > > > -------------------------------------
    > > > > 2.0 0.068 0.081 0.013
    > > > > 2.5 0.075 0.094 0.019
    > > > > 3.0 0.080 0.104 0.024
    > > > >
    > > > > All mean A_B values are in magnitudes. Basically what happens is that
    > > > > the requirement that E(B-V) < 2*(sigma E(B-V)) lets in SNe which are
    > > > > more reddened than E(B-V) < 0.1. This effect is larger at high redshift
    > > > > because the errors are larger. Note also that the average is not zero,
    > > > > and in fact the mode is not zero either. For our current method, I
    > > > > estimate that the bias is about 0.024 magnitudes in the sense that the
    > > > > high-redshift SNe are dimmer. This depresses Omega_M by a comparable
    > > > > amount.
    > > > >
    > > > > Note that if I do not include the flux-limit suppression, our bias is
    > > > > worse. So, we would expect the low-extinction technique to perform
    > > > > worse in a volume-limited sample.
    > > > >
    > > > > - Greg
    > > > >
    > > > >
    > > >
    > > > --
    > > > ___________________________________________________________________
    > > > Ariel Goobar (www.physto.se/~ariel)
    > > > Department of Physics, Stockholm University
    > > > AlbaNova University Center, SE-106 91 Stockholm, SWEDEN
    > > > tel: +46 8 55378659 fax: +46 8 55378601
    > > >
    > >
    > >
    >
    > --
    > ___________________________________________________________________
    > Ariel Goobar (www.physto.se/~ariel)
    > Department of Physics, Stockholm University
    > AlbaNova University Center, SE-106 91 Stockholm, SWEDEN
    > tel: +46 8 55378659 fax: +46 8 55378601
    >



    This archive was generated by hypermail 2.1.4 : Fri May 23 2003 - 10:27:58 PDT