From: Chris Lidman (clidman@eso.org)
Date: Wed Feb 23 2005 - 07:48:36 PST
Hi Gabriele,
I've given some more thought about the possibility of doing a 2D KS
test with the data in figure 5. It seems to me that we cannot do
this test because one is not sampling the nearby and
distant samples randomly.
The alternative is the chi-sq test, as you have done. The
resulting probability (0.15) is low. The question we must now ask
ourselves is do we believe this result. If we do, then we cannot
say that there is no evidence for evolution, as we do in the abstract.
There could be reasons for the difference. The example you give
is systematic errors in host galaxy subtraction. If this is the case,
why would if affect the the Mg II index but not the FeII index.
Regards, Chris.
On Tue, 2005-02-22 at 20:56, Chris Lidman wrote:
> Hi Gabriele,
> Thanks for the reply. I forwarded your paper to Andy and he has
> provided some additional comments. Andy's comments consist of
> grammatical corrections, which I reproduce below, and an important
> question regarding host galaxy subtraction.
>
> Before I list Andy's comments, I'd like to add to my point concerning
> the significance of the difference seen in figure 5.
>
> The test you have done is to test whether high redshift SNe are more
> similar to low redshift normal SNe or to low redshift underluminous SNe.
> You have found the former to be true. However, this does not mean that
> they are the same as low redshift normal supernova. Hence, you cannot
> conclude that there is no evolution. For this, you need to conduct a
> more sophisticated test.
>
> It may not be easy to do this in 2 dimensions. In 1D, the
> obvious tests would be the KS test and Student's t-test.
> However, using a web search, I found the following web page
> which describes a 2D KS test.
>
> http://beowulf.lcs.mit.edu/18.337-2002/projects-2002/ianchan/KS2D/Project%20Page.html
>
> Have a look and let me know what you think.
>
> In addition to Andy's comment regarding host galaxy subtraction, one
> should note that the bias could go either way. In some cases, you might
> subtract too much host and in other cases you might subtract too little.
> Currently, the paper assumes that the error in the subtraction is 10%.
> Where does this percentage come from? I think that this needs more
> justification.
>
> Cheers, Chris.
>
> Start of Andy's comments
> ========================
>
> Comment #1
>
> The largest source of uncertainty is the host galaxy subtraction. Yet
> this is swept under the rug. If this was done using the determination
> from my program, for example, you can't trust any of the conclusions
> from the paper. My program simultaneously fits host galaxy and SN
> templates. In other words, it *assumes* that the high-z spectra is
> similar to a low-z one and subtracts an amount of host light such that
> the high-z and low-z spectra will look similar. The proper thing to do
> would be to use the photometry (e.g. percent increase) to constrain the
> amount of host galaxy light, and to use imaging and spectral features to
> constrain the host characteristics. You might be able to say some
> spectrum has 30% host +/- 10%, and is either an E, SO, or Sa. Then you
> could do different subtractions in that parameter space and test the
> results. Defining a technique like this would be a great addition to
> this paper because any time people use high-z spectra they are going to
> have to do something like this.
>
> Comment #2
> ==========
>
> One other point -- in the SNLS we have many 91T/99aa SNe at high
> redshift. So there is no need to make a big deal out of having one in
> the sample, as this paper does.
>
> Minor comments
> ==============
>
> p. 1
> Intro, par. 1 -- plural of supernova is supernovae
> tidly ????
> par. 4 it should be pointed out that the effect metallicity has
> on the spectrum pointed out by Hoeflich and Lentz are only for
> early-time
> spectra, of which there aren't many here.
>
> par. 6 -- I have submitted a paper on SNLS gemini spectra -- Howell et
> al.
> 2005, ApJ, submitted
>
> p. 2, section 2.1 last sentence, One SNe -> One supernova
>
> last paragraph of section -- two days is no longer the uncertainty for
> the
> spectroscopic epoch determination in Hook et al.
>
> section 2.3, par 1 -- you say EWS but it is not defined yet.
>
> p. 4
> table 2 'Si' should be 'Si 4000' since you say it is absent in
> underluminous SNe.
>
> first paragraph on page - there is a period missing before SN 2002fd
>
> p. 5
> 3.1, 3rd par. "uncertainty in the redshifts is taken to be 300 km/s if
> the
> redshift was estimated from galaxy lines..."
> shouldn't this be proportional to the redshift?
>
> 3.2 first sentence -- delete ", they" and second comma.
>
> p. 6
> "possible involved" delete involved
>
> p. 7 last par in 3.2.1 delete space in "spectra ,"
>
> 3.2.3, 4th par -- delete space: "( with"
>
> p. 8 3.3 miss-identification -> misidentification
> this whole parenthetical expression should be made into a new sentence
> because it is confusing with "uncertainties" written twice in a row.
>
>
>
> On Tue, 2005-02-22 at 05:24, Gabriele Garavini wrote:
> > I'm very sorry chris I sent by mistake the wrong file.
> > Here is the one I meant to send.
> > Thanks
> > Gabriele
> >
> > On Mon, 21 Feb 2005, Chris Lidman wrote:
> >
> > > Hi Gabriele,
> > > Could you resend the attachment. The one I received seems to be
> > > chopped at the end.
> > >
> > > Regards,
> > >
> > > Chris.
> > >
> > > On Mon, 2005-02-21 at 13:29, Gabriele Garavini wrote:
> > > > Dear Chris
> > > >
> > > > thanks you very much for the careful reading of the paper. I've
> > > > implements most of you comments. Please find all my answers in the
> > > > attached file. I also changed extensively the section about EW (3.2) I
> > > > hope this will allow us to go to submission shortly.
> > > >
> > > > Thanks again
> > > > Hasta la vista.
> > > > Cheers
> > > > Gabriele
> > >
> > >
> > >
>
>
This archive was generated by hypermail 2.1.4 : Wed Feb 23 2005 - 07:49:27 PST