more paper comments

From: Alex Conley (aconley@panisse.lbl.gov)
Date: Tue Jan 13 2004 - 17:08:22 PST

  • Next message: VAFadeyev@lbl.gov: "I-band paper comments"

    Hello Serena,

      I hope you didn't spend too much time worring about the paper over the
    holidays!

      Here are a few more paper related comments. These comments refer to the
    December 17th, 2003 version of the paper in Letter format, and are split
    into two sections. The first section consists of 'real' comments about
    the data analysis and scientific content of the paper, and the second is
    mostly restricted to wording, spelling, and punctuation.

    Data analysis comments:

    page 2, column 1 : Maybe this is a bit late for this, or you have other
     reasons for excluding them, but there are a number of additional SNe that
     you could include that have very good I band lightcurves. For example,
     SN1999ee has incredible coverage in many bands, and SN 1998de has good
     coverage and an even lower stretch than sn1991bg. I could try to give
     you lightcurves for some of these if you would like. The worry is that when
     you start introducing a bunch of one-paper-only SNe, they aren't as
     uniformly reduced as the Calan/Tololo or CfA samples, but you've already
     started down this path with 94d, 89b, and 91bg, so you might as well go
     further and grab a few more.

    page 2, column 2 : I'm not sure I buy your estimate of 0.06 for the errors
     in the Leibengut data points for sn1991bg. It probably doesn't really
     matter, but I would assign the 0.06 as some sort of systematic error on
     all points and then try to consider the scatter of the points around some
     template fit as the statisitical error. Even ignoring this, the 0.06
     estimate probably needs to find a way to take the error bar on the
     Filippenko point that overlaps into account.

    page 4, column 1: How much would it hurt you to just dump sn1997br and
      sn1998ab? You showed that there is a cut on the chisquare that handles
      them, but I don't see how you can apply this to your real data. If you
      want to keep them I think you need to artificially inflate your error
      bars to match your uncut simulations, since this is not a cut you really
      apply. But really, it seems safest to just dump them.

    page 6, column 1: I don't understand why you excluded SN 1995E. Sure it's
     heavily extincted, but you are applying extinction corrections. Do you
     not have a good estimate for it's extinction, or do you just not trust
     the Phillips type methods to deal with such extreme cases? Either way, I
     think you need to state your reasons for excluding it more clearly.

    page 7, column 1, second paragraph:
     The 'possible correlation' needs to be quantified, at least a little,
     especially because further down you mention statistical significance
     when talking about possible other correlations.

    page 8, column 2, third paragraph:
     I think it is important that you fit the B band stretch of sn2000fr
     using K-corrections based on your spectral template, since this is what
     you use for all other fits.

    page 11, column 1:
     In addition to not applying the width-luminosity correction, you might
     also consider assigning sn1999q using the mean stretch with error bars
     determined from the distribution of B band stretches at low redshift.
     Alternatively, if Riess et al. give dm15 you can try estimating a
     transformation between dm15 and stretch and then applying that to their
     value. You might be able to put a constraint on stretch from the colors,
     but this could be very dangerous because of possible host galaxy
     extinction.

    page 12, column 2 :
      Do you really think that future homogenous data sets are going to solve
     these problems? It's going to take a bit more work than that to solve
     the K-correction discrepencies.

    Niggling comments:

    page 1, column 2, paragraph 1: Sentences probably shouldn't begin with
     E.g. Instead, eliminitate the E.g. and just start with For.

    page 1, column 2, paragraph 2: Eliminate the comma in "second peak, 15-30
     days after the initial one."

    page 1, column 2, paragraph 2: It should be SN not SNe in "intrinsic
     luminosity of the Type Ia SNe, appearing " since you are referring to a
     singular supernova.

    page 2, column 2, paragraph 1: Change 'parameters given in the table are
     transformed in magnitudes' to transformed to magnitudes.

    page 2, column 2, paragraph 3: There should be no colon between
     CfA2 data-set and SN 1997cn. You already used a : to start the list, so
     there shouldn't be any others until the list is done. A comma would
     work in it's place.

    page 3, column 1, paragraph 1: There should be no ; between SN1991T and
     (Li et al., 2001, ... ).

    page 3, column 1, paragraph 1: spectroscopically normals should be
     'spectroscopically normal SNe Ia'.

    page 3, column 1, paragraph 2: w.r.t needs another trailing . -- w.r.t.

    page 3, column 1, paragraph 3: 'In particular we note the case of SN 1994D
     that has a chisquare/dof ~ 26.' would be better as 'In particular we note
     that SN 1994D has a chisquare/dof ~ 26.' You might also want to refer
     the reader to the appropriate figure so they can see the residuals for
     themselves.

    page 3, column 1, first paragraph of section 2.3:
      'time coverage in the I-band lightcurves' should just be 'time
      coverage in the I-band'.

    page 3, column 1, first paragraph of section 2.3:
     'However, on a total of 42 supernovae' should be
     'However, of a total of 42 supernovae' although
     'However, of 42 supernovae' is just as good and shorter

    page 3, column 2, SN1997br and sn1998 ab
      This is a bit muddled. How about:
      
      -SN 1997br: In about 3% of the cases we found chisquare/dof >= 7.
       This caused the distribution of best fit parameters to have a
       different standard deviation that what the fitting procedure
       estimated the uncertaintly to be for an individual fit. Excluding
       these cases results in the expected distribution.

      -SN 1998ab: The simulated data sets result in two populations of
       parameters, which reduces to one when a cut of chisquare/dof <= 6
       is imposed. Note that approximately 78% of the simulations satisfy
       this condition.

    page 4, column 2: You should mention that your width-luminosity relation
     is between the first I band lightcurve peak and the B stretch, not the
     B band maximum and the stretch. Coming right after you describe fitting
     the B band data this could be confusing to readers.

    page 4, column 2: Have you defined z_{CMB}? It's pretty obvious, but
     probably worth describing explicitly.

    page 4, column 2: ' where the distance to each SN was calculated from its
     redshift and assuming a value for the Hubble constant, ..., corrections
     for Milky ...' should be: ' where the distance to each SN was calculated
     from its redshift assuming a value for the Hubble constant, ... .
     Corrections for Milky Way and host galaxy ...' -- that is, break it into
     two sentences and drop the and before 'assuming a value'.

    page 5, column 1: What does 'spread' mean in the 'The spread measured
     in...'? Is it the RMS around the best fit line? Nobody ever seems to
     explain clearly what terms like spread or dispersion mean in this field,
     and it drives me nuts.

    page 5, column 1: Have you defined alpha_{I}? I think you should work
     it into your previous equation.

    page 5, column 2: Again, what is dispersion? Is it the RMS around the
     line? In any case, it should be 'The dispersion measured for the data
     around the fitted line is'.

    page 6, column 1: 'weaker' instead of 'more feeble', or maybe just 'weak'.
      The undefined term spread shows up in this sentence as well.

    page 7, column 1, first paragraph of section 3:
     There should be no comma between 'considered here' and 'are in the
     Hubble'

    page 7, column 1, first paragraph of section 3:
     Instead of 'The whole redshift range spans up to 0.1.' consider
     'The maximum redshift in this sample is 0.1.'

    page 8, column 1, top :
     represent -> represents

    page 8, column 1, top :
     put 'with' between M_{I} = <whatever> and the chisquare:
     'The value fitted is M_{I} = <whatever>, with chisquare = ...'

    page 8, column 1 :
      in 'Hubble diagram for the three data set seperately' : set -> sets

    page 8, column 1 :
      Instead of saying the dispersion was measure in a sample, you should say
      the dispersion was measured for a sample.

    page 8, column 1, last sentence in section 3:
      I'm not sure the i.e. is appropriate in 'used, i.e. about 900 A redder'.

    Tables 5,6,7: It should be Epoch, not Epochs in the column headers.
      Each data point only corresponds to one epoch.

    page 10, column 2, second paragraph: The word 'being' is a bit awkward
     as used. Instead, I would recommend 'We performed a one parameter fit
     on the amplitude of maximum I_{max}.

    page 10, column 2, paragraph 4:
     Remove 'of the' from 'was selected for each of the simulation.'

    page 13, column 1, paragraph 4 :
     You define an abbreviation for LS, but I don't see where it is ever used
     again.

    page 14, column 1, first paragraph section 6.1 :
     Too passive. "Three data points are and inadequate sample to perform a
     robust analysis".

    page 15, column 1:
     "even prior the lightcurve is fitted" -> "even prior to the lightcurve
     fit"

    Alex



    This archive was generated by hypermail 2.1.4 : Tue Jan 13 2004 - 17:08:28 PST