Re: slides for today

From: Alexander Conley (ajconley@lbl.gov)
Date: Wed Mar 09 2005 - 14:36:36 PST

  • Next message: Reynald Pain: "Re: slides for today"

    Hi Reynald,

       Thanks for your comments. Sorry it's taken so long to respond;
    I have been busy applying for a Canadian Work Permit and the like.

    I have put a new version on the webpage (version 1.52) reflecting
    your suggestions. See below for details -- including some questions
    of clarification that I have for you.

    Alex

    > The sentence "disagree .. at 1.7 sigma" sounds odd : disagree is a
    > strong statement , 1.7 sigma is not much.
    > also the "favors ... at 1.6 sigma"
    > do the sigmas even include (known) systematics ?
    >
    > for the 2nd statement I would say the CMAGIC and peak mag agree within
    > less than 2 sigma (I assume that this sigma is
    > not the statistical sigma of one of the method or any combination of
    > the two since the two measurement are done with the same data. you
    > took that into account right ?

    I agree, it isn't totally obvious what to say here. Perhaps the more
    neutral
    'differs from a flat universe'. I'm not completely comfortable saying
    agrees at less than 2 sigma, since 2 sigma isn't totally negligible,
    but you are correct that disagree is a stronger statement than 1.7 sigma
    warrants.

    The numbers include systematics, and correlations, and the abstract
    now reflects this.

    > page 2 : 3rd paragraph : I think the stretch was introduced in P97

    Included

    > page 3 : 3rd paragraph : add ref Tonry03 and Barris04 who also use (I
    > think) the Bayesian approach to treat extinction ?

    Added

    > page 3 paragaph 4: you say that CMAGIC has advantages for evolution
    > but do not clearly say what the benefit is.

    Modified to say 'and possibly evolution, depending on the model.'

    > page 5 : "By studying well observed ..; " explain how

    The real details are tedious and inappropriate to include, but I have
    added the text

    'Using well observed low redshift SNe to
    determine the earliest and latest points in the linear region as a
    function
    of stretch, we find that beginning date of the linear region is ...'

    > page 9 : top : "0.1 rest frame days" why not .1 and not 0.5, 1, ...
    > ? justify specially if the analysis is very sensitive to it

    Uh, the same sentence says that 'the analysis is quite insensitive to
    this
    value'. More specifically, it can range from 0.1 to 7 without having
    any
    difference at all. Are you saying that I should simply remove this
    sentence?
    (However, that should have read 0.5 days).

    > page 17 : last paragraphe before 6.4 . i ma not sure the first 2
    > sentences are needed

    Experience suggests that people do ask how well the secret
    offset worked, so it seemed worth including.

    > systematics: Include some estimate of systematics from K-cor (other
    > than U-enhanced) (propagate uncertainties of K-corr

    I would love to have a better handle on the kcorr systematic. What do
    you
    suggest?

    If you are saying to do what Rob, Riess, etc. do, which is to add
    around 0.02
    mag uncertainty to each magnitude, this isn't a systematic. In fact,
    it's totally
    useless even as a statistical error, since it is already included
    automatically
    in \sgi .

    > Conclusion : same remark as for the abstract concerning the
    > conclusions drawn from 1.6 and 1.7 sigma deviations

    Again toned down slightly.

    > acknowledgments : I did not find 1999ee in the list of SNe you use.

    Yes... This raises an interesting point. 99ee was not present in the
    primary fit, but was used in some of the systematics tests
    (specifically,
    when the extinction correction was weakened). So I would rather not
    include it in the list of data, but I did use it and feel like I should
    thank
    those who provided it. I will add some text to the acknowledgements
    reflecting this.

    > fig 4 : what about using a smaller binning (half size or even of 0.05
    > mag)

    I played around with lots of binnings, and cutting the bins in half
    introduces a lot of noise. The highest bins get down to 8 events, so
    aren't too bad, but the flanks become hard to read.

    >
    > fig 7 : do not show the marginalized distrib ?

    I personally agree with you, but am waiting to see what other
    people think before I remove it.

    > fig 9: in the caption give the value of alpha found
    >
    Okay.



    This archive was generated by hypermail 2.1.4 : Wed Mar 09 2005 - 14:35:43 PST