From andventura@gmail.com Mon Oct 30 13:15:02 2006 Date: Mon, 30 Oct 2006 12:52:40 +0100 From: Andrea Ventura To: Stefania Spagnolo Subject: Fwd: next week ---------- Forwarded message ---------- From: Andrea Ventura Date: Oct 26, 2006 9:28 AM Subject: Re: next week To: Alan Poppleton Cc: Andrea Ventura , Michela Biglietti , Gabriella Cataldi Dear Alan, I plugged your suggested change in MuidCombined.cxx, even if not via jO, bringing the number of sigma for looseMatch from 3 to 6: the effect is to increase by 1-2% the MuidCB efficiency. Owing to problems with castor I could run on one file only (8 GeV), but in next days we'll run a larger statistics to understand better, possibly also vs. eta and phi. Clearly to (re)define this cut we need to estimate the impact on fake combined track probability (best estimated on single muons with background, even if we only have old Rome datasets). Perhaps we could study this tuning as a function of tranverse momentum. Thanks for your help. Cheers, Andrea On Tue, 24 Oct 2006, Alan Poppleton wrote: > Hi Andrea > > the SUSY problem comes from high pT InDet tracks (true hadrons) > matching to the low pT Moore fakes which often are made from very few > hits so have a large measurement error. The old loose match treated > this as catastrophic calo energy loss and accepted the match. > > The new criteria are much stricter. As the current Moore pT pulls are > too broad at low pT the matching chi2 often fails, leaving a > significant fraction of signal to be recuperated from the loose match. > To solve the SUSY problem I tightened the (loose match) energy balance > check, this needs to be slackened off to a compromise position, with > some loss of signal but nothing like as much as we now observe. > > I've spent the day wrestling with the latest nightly, for which I can't > get Muid to work at all. I think I must have checked in a bug during > the last week or so - during the time while Moore hasn't been working. > I'm now leaving for holiday so won't have a fix until next week. > > In the meantime you could play by setting the 3 sigma cut in > MuidCombined::LooseMatch to be, say, 6 sigma. I imagine that will > recover a fair number of low pT muons. That's my first guess at what a > compromise value should be. > > cheers - Alan > > On 22 Oct 2006, at 14:24, Andrea Ventura wrote: > > > Hello Alan, > > thanks for investigating on this issue. I am not aware of which are the > > problems with SUSY events: are they only at low pt or what are they > > connected to? Keep in touch, > > Cheers, Andrea > > > > > > On Sun, 22 Oct 2006, Alan Poppleton wrote: > > > >> Hi Andrea > >> > >> I looked at Michela's 8 GeV file - and see compatible numbers offline > >> (from MuidExample, selecting calo param). From a quick glance it > >> appears that I've gone from loose MuidCB cuts that were too loose to > >> one's that are now too tight. I'll carry on investigating tomorrow - > >> but we have to expect some remaining inefficency to avoid the Frank > >> Paige problems with SUSY events. > >> > >> BTW calo param is near identical to calo measurement at this low an > >> energy - since ~99% select the param f(i.e. mop value) after analysing > >> the calo signal. At 100 GeV this ratio is nearer 50/50. > >> > >> cheers - Alan > >> > >> On 21 Oct 2006, at 13:19, Andrea Ventura wrote: > >> > >>> Hello, > >>> at 11 GeV Moore efficiency wrt LVL1 is about 97% over |eta|<2.5, > >>> while > >>> MuidCombined has 91% wrt LVL1 (i.e. something like 94% wrt Moore). > >>> These numbers have been evaluated on files with only calo > >>> parametrisation. > >>> Let me know if this can be a problem. > >>> Cheers, Andrea > >>> > >>> On Sat, 21 Oct 2006, Michela Biglietti wrote: > >>> > >>>> Hello > >>>> > >>>> other two things : > >>>> > >>>> 1. I run on 1000 events, pt=11GeV. Moore reconstructs 712 tracks > >>>> (this number takes > >>>> into account L1 muon efficiency). MuidComb says : > >>>> > >>>> MuidCombinedEF INFO for 681 > >>>> tracks having a successful combined fit: > >>>> MuidCombinedEF INFO 661 > >>>> combined tracks with a good 5-param SA/ID match > >>>> MuidCombinedEF INFO 19 > >>>> combined tracks from SA with a loose MS/ID match > >>>> MuidCombinedEF INFO 4 > >>>> combined tracks recovered from MuonSpectrometer tracks without > >>>> StandAlone fit > >>>> > >>>> e.g. eff=96% wrt Moore. Is it compatible with what Andrea finds ? > >>>> The 31 tracks MuidCombined is not combining comes from bad energy > >>>> balance, > >>>> and this bring me to my second point ... > >>>> > >>>> 2. ... most of the events we are using for trigger studies are done > >>>> without calo digits. This is mandatory because we are doing large > >>>> scale > >>>> production and cannot store it. Is it possible that this is a source > >>>> (or THE > >>>> source) of Muid innfficiency? I suppose that with these files Muid > >>>> is > >>>> using > >>>> only calo parametrisation (while with calo ON one can use calo > >>>> parametrisation + > >>>> measurements). Is this an another point to be investigated ? > >>>> > >>>> Cheers, Michela > >>>> > >>>> > >>>> > >>>> -----Original Message----- > >>>> From: Alan Poppleton > >>>> Sent: Sat 10/21/2006 10:29 AM > >>>> To: Gabriella Cataldi > >>>> Cc: Andrea Ventura; Michela Biglietti > >>>> Subject: Re: next week > >>>> > >>>> Hi Gab > >>>> > >>>> I've now found 200 10 GeV muons - from Stephane's maxidisk. Still > >>>> not > >>>> enough to understand the problem. For these events I find the > >>>> following: > >>>> > >>>> 160 muons found by Moore with eta in InDet acceptance. The InDet > >>>> finds > >>>> 167 tracks => Moore eff ~ 96% > >>>> > >>>> good combined fit with good match chi2 154 => ~96% eff wrt > >>>> Moore > >>>> good combined fit with loose match 5 => ~3% (all > >>>> of > >>>> these are in the endcap, 4/5 in the CSC region) > >>>> fails comb fit with good match chi2 1 at > >>>> eta > >>>> ~1.2 on Monday I'll investigate this one some more > >>>> fails loose match (energy balance) 1 in > >>>> CSC > >>>> region again ! > >>>> > >>>> So far I have NO problems at all in the barrel, furthermore we know > >>>> there are Moore problems in the CSC region that will be reduced once > >>>> we've fully integrated David Adams' work. > >>>> > >>>> In conclusion: I need larger statistics on castor (or a maxidisk). > >>>> If > >>>> you can transfer some events, then I also need your analysis of the > >>>> efficiency for the same events. My worry now is that there's a bug > >>>> in > >>>> producing the ESD or AOD, or even in MuidStatistics, such that an > >>>> additional cut is made on say the match chi2 ?? I cc Michela in > >>>> case > >>>> she knows what we should check here. > >>>> > >>>> cheers - Alan > >>>> > >>>> On 20 Oct 2006, at 18:35, Alan Poppleton wrote: > >>>> > >>>>> Hi Gab > >>>>> > >>>>> There's a problem. The 10 GeV single muon dataset I was using has > >>>>> been > >>>>> removed from castor ! Along with everything else from the mc11 > >>>>> production - atlas management busybodies! Would it be possible to > >>>>> copy the 8 GeV data to /castor? I need around 1K events (minimum) > >>>>> to > >>>>> understand the inefficiency. > >>>>> > >>>>> I have 100 events on disk - but I don't see any problem with them. > >>>>> > >>>>> thanks - Alan > >>>>> > >>>>> On 20 Oct 2006, at 15:47, Gabriella Cataldi wrote: > >>>>> > >>>>>> at least we should think to a reasonable > >>>>>> comment to put on the slide.... > >>>>>> cheers, gab > >>>>>> > >>>>>> On Fri, 20 Oct 2006, Alan Poppleton wrote: > >>>>>> > >>>>>>> Hello again > >>>>>>> > >>>>>>> On 20 Oct 2006, at 15:37, Gabriella Cataldi wrote: > >>>>>>> > >>>>>>>> Next week is also another week and there is no > >>>>>>>> hurry... I would like to avoid people presenting > >>>>>>>> this to the triggerPhysics week with no understanding... > >>>>>>>> that is all ... there is time ... > >>>>>>> > >>>>>>> except that I'm going on vacation to Spain next week ! Hope the > >>>>>>> rain > >>>>>>> doesn't spread over there too .. > >>>>>>> > >>>>>>> However I'll be at work Monday and Tuesday - so there is still > >>>>>>> some > >>>>>>> time! > >>>>>>> > >>>>>>> have a nice w/e too - Alan > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>>> > >>>>>> > >>>>>> -- > >>>>>> Gabriella Cataldi - INFN Lecce (Research staff) > >>>>>> via per Arnesano - 73100 Lecce > >>>>>> Tel: +39 0832 297451 - Fax: +39 0832 325128 > >>>>>> -- > >>>>>> "Si, sull'orlo del baratro ha capito la cosa piu' > >>>>>> importante, che vola solo chi osa farlo."(Luis Sepulveda) > >>>>> > >>>> > >>>> > >>>> > >>>> > >>> > >> > >> > > > >