These are from the
preblessing
talk on 3-25-04. The questions are found in the minutes of the
top
meeting.
as
well as emails from Weiming, Lester and Joao.
Q from Jaco: To do the extra correction to avoid applying mistag matrix to event with heavy flavour, you need the top cross section, what is the extra uncertainty on that fraction?
A: The mistag estimate comes from applying the mistag matrix to the tagged events. The error on this mistag estimate is the statistical error from the size of the sample, an 8% systematic error due to the mistag matrix and an additional 10% error from applying the scale factor of 1.2 to account for mistags due to interactions in material. We then scale down this mistag estimate by the fraction of Wc, QCD and mistags in the inclusive tag sample. To do this scaling we simply rely on the background estimates and the number of inclusive tagged events. In this way the scaling doesn't depend on the top cross section. However, we had not been including the error on this scaling. The error on this scale factor is now included in the mistag error estimate.
Q from Un-Ki: The Pythia t-channel MC doesn't well model the 2nd b quark in the event, so the double tag efficiency will be low.
A: The single top group has generated MadEvent LO and NLO single top MC. Using this MC we find a double tag efficiency of 1.7% in the 2 jet bin for the NLO sample compared to 0.2% for the LO Madevent and Pythia samples. I combine the LO and NLO t-channel samples in the ratio of 3:1 and find a new t-channel estimateof 0.04 events in the 2 jet bin versus 0.02 for the original Pythia t-channel estimate. My total single top estimate for the 2 jet bin is 0.39 +- 0.08. This error more than covers the uncertainty due to the NLO contribution for the t-channel process. (Note that the s-channel single top events have a 10% double tag efficiency.)
Q from Mel/Young-Kee: Comment on the discrepancy in the 2 jet bin. What is the probability to observe such an excess.
A:
We observe 8 double tag events with an expectation of 3.66 +- 0.74.
This expectation is based on our background estimate plus ttbar using
a cross section of 6.7 pb (with a 20% error). The probability to
observe 8 given a mean of 3.66 is 3%. Using 3.66+0.74 = 4.4, for
the mean, we get a probability of 8%. This should not be treated as
an official number since the Poisson should be convoluted with the
gaussian uncertainty. This discrepancy by itself is not that
unusual, although we note that in Run 1, the 2 jet bin had 5 double
tags with a background +ttbar estimate of 2.41 +- 0.56. The
kinematic plots do not show any obvious discrepancy although some
have a poor KS statistic comparing data and the prediction.
Another
point of comparison is the difference between the data and the
prediction for the inclusive tag sample in the 2 jet bin. In that
case we predict 72.5 events and observe 73 tagged events. A typical
ratio for the double tag to inclusive tag efficiency is 20-25% (from
ttbar or single top). If the excess of 4 events in the 2 jet bin is
due to a new process with 2 b jets, we would expect an excess of
16-20 events in the inclusive tag sample.
Q from Lester/Weiming: The qcd estimate of zero events should have an error on that estimate. The qcd background estimate is small but zero is not the right answer.
A: The qcd background estimate comes from using the lepton isolation and Met sidebands. The low Met non-isolated lepton region (region B) has 133 tagged events in >=2 jets, none of which are double tagged. We use the 90% U.L. of 2.44 events given 0 observed events to get a double tag to single tag qcd background ratio of 1.8%. We apply this to our estimate of the single tag QCD background estimate to get the error on our qcd estimate of 0. This is 0.22, 0.09 and 0.03 for the 2, 3 and 4 jet bins respectively. We would expect some small amount of qcd background just due to mistags of the 2nd jet. This should be at the mistag level of about 1%. This contribution of the qcd background is included in our mistag estimate and provides the central value for our qcd background contribution. The additional error derived above will be added to the total error of our background estimate.
Q from Jason: It is ok for the number to put together QCD and mistags, but as a shape they do look different.
A: The qcd contribution is extremely small. Its contribution as part of the tag+mistag background is estimated to be 0.03, 0.07 and 0.04 events for the 2, 3, and 4 jet bins respectively. This contribution is negligible for the double tags and will not be included as a separate shape for the double tag plots.
Q from Weiming: Given the excess of double tags in 2jet bin, people may wonder if the method II could be underestimated. It would be useful if you could try to estimate using the method I.
A: The tag matrix is for tagging single jets and is based on the average HF content times tagging efficiency of jets in the inclusive jet data. Since it doesn't include any correlations as to the HF content of the jets, I don't think it will give a reasonable estimate of the double tag background. A tag matrix for tagged events might provide a useful way to do a method 1 estimate, but that will need to be looked at in the future.
Q from Weiming: In your talk of page 5, I still do not understand these correction factor of mistags. It's true that the Wbb Wcc was estimated using Monte Carlo, but you need to include the real mistags from data. Or am I miss something here.
A: The Wbb, Wcc MC is used to measure the efficiency to tag HF jets for those events. In addition, I apply the mistag matrix to light quark jets in the Wbb, Wcc MC so am including the mistag contribution to those events as well.
Q from Joao: Could you provide the following plots: We are concerned about tags in the material.
1) Delta-phi between the double tagged jets.
2) phi distribution for tagged jets.
3) Location of SecVtx vertex in the r-phi plane.ditional plots requested.
A: The plots are below. With the limited sample size it is hard to draw any strong conclusions. We find that 7 of the 32 tags in the double tag sample have a radius greater than 1.3cm. This is 22% of the sample, compared to an estimate of 15% from the MC. Note that for all positive tags in the stripped electron sample (about 15K tags) we find 13% are at a radius of 1.3cm or larger. Our mistag estimate does include the correction factor of 1.2 to account for interactions in the material. The deltaphi plot between double tags has a poor KS statistic. That is because the events in the 2 jet bin are not smoothly distributed across the plot as expected.
Q from Jaco: Plot the Dr and bb mass from the 2 tags in all the bins. not using the nearest jet, but the two tagged jets.
A: The plots are below.
Q from Nigel: Are you going to emphasize the uncertainty on the background on the main Njet plot? Background uncertainty is hidden in jet multiplicity plot, suggest to put band around total shape.
A: Here is a plot with the total error on the ttbar+background prediction.
Q from Young-Kee: Give the chisquare/KS to test agreement for distributions.
A: The KS test value (using option X) are put on the plots above. Other plots shown at the preblessing with the KS value can be seen here. The values tend to be on the low side. The worst two are the Met and the deltaphi between the tagged jets. Given the 2sigma discrepancy in the observed number of tags in the 2 jet bin, the level of agreement for the plots seems reasonable.