From - Fri May 9 19:23:05 2003 Return-Path: Received: from postala.lbl.gov ([128.3.41.61]) by imapd.lbl.gov (Netscape Messaging Server 4.15) with ESMTP id HEKYXP00.43M for ; Thu, 8 May 2003 11:20:13 -0700 Received: from postala.lbl.gov (localhost [127.0.0.1]) by postala.lbl.gov (8.12.9/8.12.9) with ESMTP id h48IKAec016352 for ; Thu, 8 May 2003 11:20:12 -0700 (PDT) Received: from sierra.lbl.gov (IDENT:0CFEee9F4OIKYfMflcaAsErFFdvKWgj9@sierra.lbl.gov [131.243.48.105]) by postala.lbl.gov (8.12.9/8.12.9) with ESMTP id h48IKArU016350; Thu, 8 May 2003 11:20:10 -0700 (PDT) Received: from localhost (deg@localhost) by sierra.lbl.gov (8.11.6/8.11.2) with ESMTP id h48IKA615503; Thu, 8 May 2003 11:20:10 -0700 Date: Thu, 8 May 2003 11:20:10 -0700 (PDT) From: Don Groom To: Tony Spadafora , Saul Perlmutter Subject: The Lewin, Perlmutter, et al paper Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear Tony, Saul, and policy board members, By now I've spent considerable time going over "Sensitivity to analysis technique in cosmological results from Type Ia supernovae," by A. Lewin, S. Perlmutter, et al. It's a very interesting and important paper, and a lot of work has been done. One of my problems is that the title isn't what it's about. In addition, I think there are some presentation matters which muddy and confuse things. I also made my usual hundreds of texteditor's comments on my printout, but that's a separate matter. The three different approaches by the 2+ groups give different values for \Delta_{\rm corr} and, via m_B^{\rm fit}, for the distance modulus. The study and comparison is valuable, but no conclusions are drawn re the cosmological results. How about a title something like "Sensitivity to analysis technique of peak magnitude determination for Type Ia supernovae"? Along the way, the material about {\cal M, D} could go. This is in e.g. P99 anyhow, and I don't know that it has bearing on the content of the paper. Since people are still confused about where H_0 went, the introduction could simply state, as a one-sentence paragraph, something like "Since the cosmological constants are extracted from the shape of the function describing m vs log z, the analysis is insensitive to constants added to m and hence is insensitive to H_0." I. Introduction My only problem here, except for the promise that cosmology changes due to the systematic differences will be discussed, is that MLCS is introduced without definition in \P 3. 2. Light curve fitting methods The central point here is that SN width or equivalent is correlated with magnitude at maximum, the magnitude correction being \Delta_{\rm corr}. Why not describe this problem (as summarized in the last two \P's of the section) at the beginning of the section (along with distance modulus, if necessary)? Then proceed to describe the methods involving s, \Delta M_{15}, and MLCS, being as specific as possible in each case as to what \Delta_{\rm corr} is. For example, how about a plot \Delta_{\rm corr} vs s for SCP objects? Where does \alpha come from, and how big is it? It's treated as a nuisance parameter in the fits. It would be nice to show a plot of peak M vs s and/or vs \Delta M_{15}, for example, as an introduction of the plots comparing \Delta_{\rm corr} by the different methods. \alpha *is* discussed under "stretch with host extinction" in sec 5.1, but for pedagogical reasons something should be said here. Maybe I'm missing something, but I don't understand the relevance of distance modulus, except that it is a measure of the residual fitting-dependent-errors in m_B^{\rm fit} after \Delta_{corr} is subtracted. (In P99 the word "modulus" does not appear, although "distance" appears 8 times.) I agree that "concentrating on distance moduli" (1st \P, section 4) focuses on the combined differences between the methods (\Delta_{\rm corr} plus all the fitting nuances), so will not comment on this further. Not sure how to do it, but a tabular summary would clarify and contrast the methods and their differences. This section (halfway down the first page) says that all methods used data from two different filters to find extinction, etc. As is said later on, P99 did not always do this. It's a little muddy. 3. Light curve analysis This might be my most serious problem with the paper: Where do the numbers come from? The data sets are carefully described, but did the authors actually reanalyze all the data by all the methods? Starting from what? Raw images? That would be monumental, involving even for our case carefully reduced lightcurve data, templates, and K corrections for each z. We've discussed doing an SCP analysis of the High-Z data in connection with Rob's paper (and I guess Alex is doing it?), and it's not trivial. And along the way we've we've gone thru 10^5 data reductions for each SN to provide the points to which the stretched lightcurve is fitted. So what was done here? And my old saw: why even give lip service to the High-Z teams use of the biased one-sided prior? Is it relevant here? 4. Differences at low redshift. 5. Reasons for differences [presumably still restricted to low-z] All of the methods give results sensitive to possible evolution, host extinction, and other physical effects, and are sensitive in different ways to the nature of the data (early points, big errors, etc.) It's appropriate to focus most of the attention on low-z SNe because of the better measurments, as is done in the paper. Figs 1-7 are all referenced in sections 4 and 5, but the captions for Figs 4-7 don't say so. Figs 4 and 5 have the same number of points, so they must still be low-z. Harder to count in Fig 6, but probably the same. But how about fig 7? Are some high-z guys included here? The captions should say. The lower panel of Fig 6 seems a little hollow; it certainly is hollow at high-z where R has all the pull. Is there any content here at low z? Why not add the diagonal straight line in Fig 1? And in the other figures some of the fits discussed in the text? In Fig 7 the vertical scale should run to at least +0.10 to show the error bars near m_B = 0. Are the points taken from the literature? And, as always, I have trouble with large error bars being shown as symmetric on a log intensity plot. At first sight Figs 2 and 3 are redundant, but I guess the divisions between Groups I and II is more striking in Fig 2. And the systematic change with \Delta_{\rm corr} in some of the permutations shown in Fig 3 are not so evident in Fig 3. Good! Our historical two-parameter relationship between s and \Delta M_{15} is shown several places, and in all of them a straight line would do just as well. With 7 points below the solid curve in Fig 5, a straight-line fit is arguably more appropriate. If we are to introduce the 2-parameter fit, more discussion of it in the text is needed. One of the most striking results of the paper is the double-valuedness of \delta^{\rm MLCS}_{\rm corr} vs \delta^{\rm stretch}_{\rm corr} in Fig 2, and the same phenomenon in Figs 2 and 3, depending on the presence/absence of pre-max points. This shows up as well in \delta^{\rm Dm15}_{\rm corr} vs \delta^{\rm stretch}_{\rm corr}, but the effect is not as big. This is all worrisome. I think one can dismiss MLCS as not being as robust as the other methods (besides being rather stupid), but this is a very real systematic difference between the methods. This is all discussed very well in the "Observations before maximum light" discussion. Perhaps the generally greater scatter of MLCS is not emphasized enough. But in this discussion it is assumed that the stretch method "does it right." Is it possible that the physical content is in the shape of the post-max lightcurve and that it is the s guys who are biased? I studied this in some stillborn work in connection with the GersonPaper, and concluded that s *can* be extended to early times. But in the context of this paper the implicit assumption that the s method is better is hard to justify. The missing plots are the \Delta_{\rm corr}'s vs distance modulus. This would convey very important information not easily visible elsewhere. This is a rather chaotic review, but I found the paper, particularly sections 2 and 3, rather chaotic. It is an impressive analysis which deserves to become more focused. Throughout, tables might be useful, if only for pedagogical reasons. Lots of fits have been done and are only mentioned in the text. And, please, to make things easier for a reviewer, write \parskip3pt \parindent0.5in at the beginning! And at some point spellcheck. I also wrote down lots of suggested rewordings for sentences. Hope this helps. It's meant to be constructive! Don |-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+|-+| Don Groom (Particle Data Group, Supernova Cosmology Project) DEGroom(at)lbl.gov www-ccd.lbl.gov Voice: 510/486-6788 FAX: 510/486-4799 Analog: 50-308//Berkeley Lab//Berkeley, CA 94720