After the application of the necessary K-factors to make the Alpgen inclusive W and Z cross sections correspond to the ones measured at CDF (see here), we obtain the following for the Z,
In Muons
Total numbers of events: N(BG)=5598, N(data)=6208
Continue reading W/Z Backgrounds with Alpgen K-factor and Comparison to Pythia
Vadim pointed out to me today that the Alpgen cross sections for W and Z do not agree with the measured CDF values. From J. Phys. G: Nucl. Part. Phys. (2007) 2457-2544 we have the measured values:
where the Z is measured for 66<116 GeV. The corresponding Alpgen cross sections are
where the Z is in the range 75<105 GeV. Ignoring the difference in dilepton mass ranges for now, we obtain the following K-factors:
The authors of the top dilepton cross section apply similar K-factors, as can be seen from comparing Table 6 in CDF 9400 with the top group MC webpage.
–Dan
Carla, Scott, and I discussed what steps are needed to complete this analysis. We have come up with the following:
If others have more steps or details to add, please leave comments on this post.
–Dan
I have added PV information to the ntuple and modified the selection of Zs. The selection is as follows:
I have added diboson backgrounds. With these selections I get n(Data)=6208 and n(BG)=4020. Plots below the break. Continue reading Z with improved selection
Here are the final soft muon efficiency / fake rate plots. I am not comparing these numbers to the ones found by the B group, since I’m using different definitions than they did: I’m counting muon candidates by tracks instead of by muon objects. The B group was finding a muon object, checking whether its track could be identified as a muon or hadron, and then applying the likelihood. I’m checking all tracks to identify them as a muon or hadron. If a track is identifiable by a secondary vertex fit, but it doesn’t have a muon object, I give it a likelihood of -2. Since most hadrons don’t leave a muon object in the detector, I get much better numbers than the B-group did.
As a first basic check of our background modelling, alpgen cross sections, whether I have made a massive mistake, etc., I look at inclusive Z’s and W’s in the CMUP (CMUP-CMUP in the case of Z’s) sample. I do not have the QCD contribution included, but it should be negligible for the muons. I also do not have dibosons, but their cross section is small. I expect agreement in the ~5% range. Instead I see the following,
For W’s with the selection pT(CMUP)>20 GeV + MET(L5)>20 GeV : n(BG)= 238974, n(data)=425875. Below is the corresponding transverse mass plot.
We are clearly missing a lot.
For Z’s with the selection 2 CMUP with pT>20 GeV: n(BG)=4011 and n(Data)=6224. Below is the dimuon invariant mass,
I checked the njet distribution and the slope of the background and data are similar, but the data is high everywhere. After some investigations, I see that I may not be selecting the primary vertex correctly and also that I am not saving all the PV information in my ntuple that I should be (quite stupid of me actually). The next thing I need to do is reprocess all the ntuples to put this stuff in there. I am beginning this job now.
–Dan
Here are the plots of efficiency vs fake rate for the soft muon code, separated by muon type. The efficiencies and fake rates are similar to those quoted in CDF Note 7043, although the note did not give the Pt dependence. I apologize for the low statistics in some of these categories. Plots are below the fold.
Update 6/8: Cleaner-looking plots (see bottom of post)
-Scott
I defined a control region similar to the one above as follows:
I used L5 corrected met because I want to plot the dPhi between the met and clusters/leptons and it will be badly smeared if I don’t correct for jets. I only look at the muons because I expect them to be have less QCD background. The background prediction (excluding Z+jets, dibosons, and QCD) is 4933 events and we observe 5609 in 950 pb-1. The plots are below. Continue reading Delta Phi(Met) for muons in a control region
I inverted the cluster likelihood selection and compared the resulting distributions to first 950 pb-1 of data we can use (0d and 0i) to check our modeling. The selection for this cut set (#8) are as follows:
Following up on this post where we see that there might be some advantage to having separate likelihoods for the case where a cluster contains 1 and 2 tau, I have created separate likelihoods. I used the following variables for the likelihoods: