Minutes of the muon software meeting Jan 21 2004, taken by Victoria. 1. Michael Gold: MC stub-track matching distributions ------------------------------------------------------- Michael's aim is to tune the simulation to get the matching variables (dx, dphi etc) to agree better between data and MC. He is using DY MC which has a good match in the muon pT distribution with the high pT muon data. For pT<10 GeV, dX matches ok between data and simulation, for 1020 GeV there are tails in the data which does not appear in the MC. The residual distributions look strange for the CMX. This may be because the drift model used to calculate the residuals may not be the same as that used in the reconstruction. After some discussion we decided that it would be a good idea to change the MuonSignedHit::getResidual() method so that the drift model must be passed as a argument. (A second option, which is to add information about the drift model to the Stubs is probably too late and too complicated to implement.) There are two options to get the matching variables to match better: 1. Smear the hits, which will smear the stub position. 2. Smear the resulting dX distributions. Everyone agreed that the first option is the preferable one. Michael is going to try this first. If it's too hard to implement, he will then try the second option. Michael hopes to have some first results by next week, and definitely in time for the next muon meeting. 2. Gavril Giurgiu: Likelihood-Based Soft Muon Tagger -------------------------------------------------------- Gavril presented some work he is doing with Min-Jeong Kim, Vivek Tiwari, Manfred Paulini and James Russ to build a soft muon tagger. In this analysis J/psi->mu+mu- events are used to select muons. Ks->pi+pi- events with a stub in the detector are used to selection fake muon candidates. Some of this "fake" sample contains muon decay-in-flight events, which are genuine muons, but we'd like to reject them from analysis as background anyway, so that's ok. The Eem, Ehad, dX and dZ distributions for the real muons are parameterized by functions. These shapes can then be used as probability density functions to determine if a particle is or is not a muon. Gavril found that the dX distribution, and dZ for CMX, is strongly dependent on pT. Ehad also shows a mild dependence on pT, which Gavril found a little worrying, but after some discussion we agreed that this was not an unexpected effect. This pT dependence is taken into account in the tagger. This is work in progress, but it already found that using this method of tagging is an improvement compared to making a square cuts on the variables considered. More improvements will come with optimization. Comments: Lucio said that he like to move toward having one unified muon tagger, integrating the work on the SLT with some of the work presented here. Lucio asked if they had looked at the dPhi distributions. Gavril said that he thought they were highly correlated with the dX distributions. Lucio said that they were, but he found that even so they were very powerful for separating muons and fakes. Lucio also said that correlations between the variables used here should be taken into account. 3. Ken Bloom: Muon ID efficiency vs. dR(TL-jet) ------------------------------------------------ Ken has been worrying about the observation that the muon ID efficiency scale factor seems to drop as a function of the number of jets in an event. (The scale factor is the ratio between the efficiency since in the data and in the MC.) The statistics used in this observation are small. Using Z+jet events Ken looked instead at the efficiency scale factor as a function of the distance between the lepton and the jet. This is a measure of how a nearby jet effects the muon ID efficiency. In answer to Michael Gold question, Ken said that the isolation cut is included in the ID cuts. The jets being used in this study have pT>15 GeV. The jet must be at least R=0.4 from the lepton. The efficiency scale factor is flat, within statistics, for both CMUP and CMX muons. To increase statistics the results for pT>8 GeV jets are compared to the pT>15 GeV jets. The efficiency scale factor for both of these types of jets. In order to see if there is any dependence on the total number of jets in the event the scale factors (using pT>8 GeV jets) were compared for events with >=1, >=2 and >=3 jets. No significant difference in these three samples was found. In conclusion, from this study there is no evidence that the scale factor of the efficiency drops with the number of jets. So we can all stop worrying! A generous systematic error of 5% is going to be used to cover any residual effect from the number of jets in an event.