Malmquist bias section

From: Greg Aldering (aldering@panisse.lbl.gov)
Date: Fri May 02 2003 - 09:39:08 PDT

  • Next message: Isobel Hook: "HST paper - Isobel's commenst part #2"

    First, so Rob doesn't miss it:

    IT LOOKS LIKE THE 0.01 mag IS MISSING - I THOUGHT THIS WAS STATED.
    WE HAVE 0.02 mag AT HIGH-Z and 0.03 mag AT LOW-Z. THE DIFFERENCE IS
    0.01 mag, AND SHOULD BE STATED IN THE LAST PARAGRAPH OF 5.5!

    Now to continue ...

    Since the Malmquist bias section has drawn some questions, let me try
    to address them and then we can see whether there is a better way to
    say what little can be said.

    First, realize that the standard Malmquist bias correction is applied
    to the mean luminosity of an ensemble of standard candles. It is not
    applied differentially based on the distance of the object, although
    that is how the bias acts - differentially. There is a standard formula
    for ordinary standard candles that relates the amount of bias to the
    power-law slope of the counts of objects versus flux and the disperion
    in the standard candle. Intuitively you can see that if the standard
    candle has no dispersion, there can be no Malmquist bias. Likewise, the
    bias is less if there are few objects in the parent population near
    your flux limit because this implies that you have found all the
    objects. If you find all the objects there can be no Malmquist
    bias. In P99 I determined that the SN sample had a number-vs-flux
    distribution consistent with that of the galaxy population as a whole,
    and so used that as the power-law slope.

    Next, it is important to realize how SNe are different than other
    standard candles. Since SNe brighten and fade, and occur at random
    epochs, it is possible that the discovery criterion is completely
    uncorrelated with the peak brightness of the SN. In this case there is
    no Malmquist bias because you have not selected against underluminous
    (for a given stretch) objects. In P99 I used the formula from Jeff
    Willick's paper to include this de-correlation suppression of Malmquist
    bias. In this paper I did not use it, since most of the SNe were
    discovered at maximum light, and hence their selection is strongly
    correlated with their peak brightnesses.

    Now, I will be the first to admit that the above approach is
    approximate. What would be needed to do it right? Well, we would need
    to know the flux-limit for each search field and determine which SNe
    could have been found, using a full Monte Carlo of SNe brightening and
    fading. (For P99 the flux limits were known - they have not been
    determined for the current searches.) We would have account for the
    fact that different scanners searched with different thresholds (which
    we do not know because these are not saved). Next we would have to
    account for the selection of candidates for spectroscopic follow-up,
    which is again subjective. Next we have to simulate the spectroscopic
    follow-up to determine whether we tend to get the best spectra of
    overluminous SNe, and know how the host brightness and SN epoch
    interact when doing spectral classification. Since we rarely get
    spectra of all the candidates (this was especially true of the
    March/April 1998 search), we have to not only know the cut-off for
    candidates, but all the details of why one or another SN was chosen to
    be observed over another. Finally, we have to account for the fact
    that we wanted to send the highest redshift SNe to HST, but that we
    also had a fixed number per region of sky that we could send.

    Roll that all up into a Monte Carlo, and you're done! So, I hope you
    can see such a job is impossible for the actual conditions of the SNe
    we are reporting. This is much different than running a nice clean SNAP
    simulation and then looking at Malmquist bias. I tried to get some
    estimate in the face of all this. I tried not to worry the reader too
    much, assuring them that the net effect was small, i.e. 0.01 mag.

    OH, IT LOOKS LIKE THE 0.01 mag IS MISSING - I THOUGHT THIS WAS STATED
    WE HAVE 0.02 mag AT HIGH-Z and 0.03 mag AT LOW-Z. THE DIFFERENCE IS
    0.01 mag, AND SHOULD BE STATED IN THE LAST PARAGRAPH OF 5.5!

    As a practical matter, I am not wedded to the paragraph starting
    with "Since Malmquist bias ...". I wrote that isolation and didn't know
    what tone the other sections would take. If there is a reason to
    keep that paragraph, then a simple fix is to say "... to warrant
    more detailed modeling of this effect than presented here." Next,
    one could eliminate the clause ", nor are the constraints on the dark
    energy equation of state suffciently strong," (although that statement
    is still true).



    This archive was generated by hypermail 2.1.4 : Fri May 02 2003 - 09:39:09 PDT