Here are the replies to your latest comments on the gamma-b paper. Ray >From FRASCATI@axcdfq.fnal.gov Mon Apr 9 19:05:51 2001 >Date: Fri, 23 Mar 2001 16:19:56 -0600 >From: FRASCATI@axcdfq.fnal.gov >To: velev@fnal.gov, v0565@fnal.gov, rebcdf@fnal.gov, goshaw@fnal.gov, > bed@fnal.gov, rlc@fnal.gov >Subject: 5450 q&a > >From: SMTP%"rlc@fnal.gov" 20-MAR-2001 18:02:10.80 >To: FRASCATI >CC: >Subj: Re: comments on cdf 5450 > > dear cdf 5450 godparents, > I don't agree that most of my questions have been answered > in a satisfactory way. A few examples: > > 1) the gamma+b analysis compares data to a method II prediction > of the SM background. This prediction has an error of about 16%, > close to the error of the method II prediction of the background > to W+b events. I was impressed and I wanted more details on how each > background and its > uncertainty were estimated. I also thought that, before claiming such an > accuracy on the background calculation, the authors should have shown > that they understand the same backgrounds in other data samples. > This is standard practice in a ``conservative and careful" collaboration. > There is nothing like this in the answers: > a) I am not told > how the rate of negative tag was parametrized. SECFAK_II from version_7_12 > b) there is no way > of understanding in the paper or in the documentation how > the rate of fake photons is determined The fake photons are determined by the CES and CPR methods. The concepts of the methods are explained in the paper, but we do not attempt to discuss them in detail or to prove their effectiveness. We did not think this was necessary since the methods have been used several times before and have been documented and published and we reference those publications. With such a long paper trail we chose not to repeat the information in our notes and papers. > c) the authors believe that they tag jets correctly, but a > tautological answer is not good enough. Good answers are > "Weiming checked it" or " here is the list of tags, check it". The tagged events are in cdfsga:/spool/rlc/ghib_tag2.pad Reproducing the tags from the inclusive photon sample will be harder since we would have to go back to the tapes. This file should be all events that pass the phton cuts and contain a secvtx tag > d) I am glad to know that Keith (who should then be referenced > in the PRD) and not MLM calculated the > matrix element of gamma b production. > Nevertheless, I would like to > see some description of its implementation into Herwig > (and some checks too). I remain > that the uncertainty of the calculation is not > discussed seriously (remember that the Ellis calculation > of the bbar production is off by a factor of two). And we include a factor of two uncertainty. Since this is a relatively small part of the background, increasing the uncertainties just doesn't make much difference. For the samples with met recall we include another 100% systematic uncertainty. We added the references. > >2) I thought that showing a comparison of data and prediction up to > a jet multiplicity of 8 was silly. Your answer: > "The higher jet multiplicities would be predicted by the standard > Herwig; they were not specifically checked." > We have removed the background for more than 4 jets in the sample without met and more than 3 in the sample with met. >3) There is no point in showing Figure 33 with ``a posteriori probabilities". > The problem is not the ``a posteriori" as you seem to suggest, but the > fact that a reliable background calculation is missing. You say > this innn the text, but a confession is not an absolution. > In addition, one is not > supposed to sin and repent in the same article. > >4) There is no point in showing double tags without an evaluation > of the background We do not agree that every event count requires a background estimate. Some numbers can serve to simply describe the data and archive the results. We would prefer to report items that we can, given reasonable time constraints. > >5) There is no point in singling out events by run and number when there > is no solid background estimate or no excess. > We disagree on the point of providing event numbers, it is a record of the results and easy reference - we simply don't see the harm. > I understand that, since there is not a SUSY discovery, the attitude is > very cavalier and errors and procedures are let to be handwaved. > It makes no harm here, but helps diffusing > in the scientific community the perception that we are a conservative but > not careful collaboration. On one side, I find it regretful that you are > sacrifying an interesting QCD measurement to the SUSY hype. In Run II, we hope that the signature-based search technique will de-emphasize the models and promote studying the background. We think the new searches will look more like QCD studies. > On the other hand, > imagine that actually CDF (or D0) will > find out in Run II that the all calculation is wrong > by a factor of two and there is actually a huge excess. How will you feel > having set a 16% error now? We all take a risk of being wrong when we publish. We are completely satisfied that we have been conservative in assigning systematic uncertainties such that there is very little chance of serious error. > In addition, all this showing of events with MET and double tags will > backfire when taken > out of the context of the determination of the background, > where they belong. > > regards, Paolo > Ray Culbertson CDF/FNAL rlc@fnal.gov 630-840-6744 Trailers 170F (MS318)