From: Robert A. Knop Jr. (robert.a.knop@vanderbilt.edu)
Date: Sat Jul 12 2003 - 06:41:38 PDT
On Sat, Jul 12, 2003 at 03:14:04AM -0700, Saul Perlmutter wrote:
> Rob, it took me a rather long time to figure out why your residual
> table didn't match the plots in the paper. I eventually translated
> (most of) the names in the table to the IAU names, and figured out
> that the following SNe are in the residual table you sent but not in
> the paper or the other master tables you sent: 975, 9617, 972, 9739,
> and 9748 (which are 97J, 96ck, 97G, 97S, and 97K). Do you know what
> these SNe are?
Those are the supernovae with really horrible color errors, which were
not even considered part of the "complete set" for the paper. I had
thought that they weren't even included in the full set I was running
residuals for, which is why they didn't get tagged as not a member of
anything. They should be thrown out, as they were from the summary
table posted online.
I'm sorry I included those. I didn't think to check for it because I
thought I'd trimmed down that table to the point that they were no
longer hanging around anywhere. It's been a few weeks, though, so
clearly I'd forgotten where I'd had things.
> Also, as you can see in the fourth plot in the attached Postscript
> file, there still seem to be some rather small discrepencies between
> where the points appear and where they are in your plot of Figure 5.
> Did you send the actual data used in the paper?
Not exactly. As I said, there will be some roundoff errors. The actual
data used in the paper is the summary table as posted online. The data
I used to generate these residuals comes from that table. I didn't use
residuals myself when generating the Hubble diagrams, and I don't
understand why you need them to do that.
Ah-- one other difference. It seems that for plots in the paper, I used
a slightly wrong scriptm (-3.491 instead of -3.481), which will
(obviously) move all the lines by about 0.01 magnitude relative to the
points (who are plotted purely from data, and won't be affected by
parameters). If you're looking very closely, you will notice the effect
of that.
> 7. A few edits for Tables 4, 5, and 6: First, these tables need two more columns (with
> error bars) giving the m_B^effective (after stretch and K-correction, etc.) and the
> m_B^effective after color correction (but only for the primary set
> SNe).
I don't think these are necessary.
First of all, the m_B column already includes everything except for the
stretch correction. Yeah, the error bars don't have everything, we do
explicitly list in one compact place all the various uncertainties used
(which is a step up from what either group did previously-- it was there
in our paper, but took some divining to figure out). (That is, peculiar
velocity, stretch, 0.1 of galactic extinction, and intrinsic dispersion.)
The danger of including what you're looking for is that they are *not*
what you would use to reproduce cosmology fits. Specifically, because
alpha (and the alpha used for uncertainty propogation) varies in the
cosmology fits. Except for the 0.17 magnitudes of intrinsic dispersion,
the stretch error is the primary uncertainty in addition to the plain
uncertainty on m_B.
There is also the problem that the tables are getting a little too wide
as is.
> These columns will allow other people to make the same Hubble diagrams
> that we show -- and then they will be much more likely to use this
> paper's results (and cite it) in the future literature. (I hope this
> is not difficult to do -- I think it's important..) Second, is there
> any reason that there are three significant digits after the decimal
> point for many of the values *and* their error bars? This looks
> wrong, since there are no uncertainties that drop below 0.01.
Yeah, but there are a bunch that are between 0.01 and 0.03, and there it
makes a fair difference (up to 50% of the error bar at 0.01, still 16%
of the error bar at 0.03) what the third digit of the uncertainty is.
"Significant digits" is a very vague and approximate concept that is
really only appropriate when when you're doing a very approximate
estimate of your uncertainties, or when talking to students for whom
actual uncertainties would be too advanced a concept. In any event,
it's better to give a little bit too much information, since extra
irrelevant information is just meaningless and can be ignored, than give
too little information for some of the supernovae.
I guess my main gripe with all of this is that it's rather irritating to
be continuing to make all these fussy and detailed changes to this damn
paper. Was it not possible to come up with all these sorts of picky
format changes during the man weeks before we submitted it?
It's a simple fact that as long as we keep looking at it and thinking
about it, we will come up with things that we thing, gee, we could do
this a different way. Part of the process of publishing a paper is
deciding to *stop* going over and doing things differently and to just
send the thing in. Accepting that there might have been some other ways
to do things, and that they might even have been better to some ways of
looking at it, but that eventually there comes a point of diminishing
returns. I'm not talking about finding and fixing problems, which
obviously you do (although with a paper of any size, it's inevitable
that there will be small stupid things), but about these kinds of
different approach or format changes. I can't help but wonder if the
paper would still not be submitted if I hadn't been such a pain in the
ass a month and a half ago about getting the thing submitted despite
everybody else wanting to continue dicking with it. My greater fear is
that after we go through this round of table fixing and figure
inserting, that there will be a new round of things here and there, ad
infinitum.
-Rob
-- --Prof. Robert Knop Department of Physics & Astronomy, Vanderbilt University robert.a.knop@vanderbilt.edu
This archive was generated by hypermail 2.1.4 : Sat Jul 12 2003 - 06:41:43 PDT