Questions from Pre-Blessing on April 28th @ the Top Meeting
Questions
Remarks/Request
Can you look at at jets that are tagged by SLT but not by SVX?
Can you find out why these jets are not tagged by SVX?
We have released the Et cut on the tight jet definition from 15 GeV down to
13GeV (which corresponds to about 1 sigma of the jet energy scale) & then further down to 10 GeV.
The expectation of the background is scaled according to the MC.
Notice that:
moving from 15 to 13 GeV, there are 2 events that migrate from 2 to 3 TJ, hence indicating a
statistical fluctuation is likely responsible for the 3TJ being low. The expected bin migration is not
even remotely that big (it's 3-4%). Therefore, there is no problem with the ttbar, ttbar is there.
Down to 10 GeV, there actually are 3 extra events. Things are about in line with the expectation.
Because a statistical fluctuation is likely responsible for the low 3TJ, it isn't relevant to seek why SecVtx event aren't
tagged by SLT or vice-versa. They aren't because they are not suppose to be at any substantially different rate than observed.
Slide 9
Sal: Looks to me that actually the MC is predicting SLT+SecVtx better than SecVtx only.
Have you tried to take the ratio between SecVtx & HJ ?
Here are the original plots of NGoodTrk for secVtx only and SecVtx & SLT Tag using the enriched
bb sample that trigger the questions. Here are plots of the Data/MC ratio.
This was the only plot that shows that the MC does do such a great job in reproducing the number of
good SecVtx track for SecVtx tag jet. For double tag jet the MC is in good agreement with the data.
It is conceivable that the data/MC ratio is different in the two samples, but the statistics in the SVX&SLT
plot are pretty limited and it's hard to say that it disagrees with the SVX plot.
In any case, we are not very sensitive to the SVX scale factor because we are normalizing to the
number of observed SVX tags.
Jaco: This jet Et of about 25 GeV, not same Et as ttbar.
Concerned about the ability of predict the tagging efficiency in high Et jets.
How do you check the double tagging efficiency for high Et jets ?
Concerned related to the ability of predicting the ttbar in >=3 jets bin.
Here are few slides to answer the concern.
Slide 24
Jason: The 95% CL looks odd in >=3 jets.
The limits given are for additional non-SM signal past the ttbar+bkgd expectations.
I find it hard to believe that a model predicting 3.2 evts in addition to 4.7 ttbar+bkgd
(giving a total of 8 evts) is not excluded @ 95 percent CL by observation of only 1 evt.
The answer is in the fact that this is a Bayesian calculation, and therefore it is the observation that
counts most (1 event), not what one "might" have observed (4 events). The Bayesian approach takes
to face value the observation as the SM background (the expectation may have fluctuated down to 1 event)
and tells how many events are allowed as extra signal. In other words the limit is ~independent of
the expected background. The method is pretty standard and it is what is suggested in cdfnote 7117, (Sep04)
by the statistics committee. One can check Table 1 on page 6 and the discussion on page 5 and page 10
(2nd par), which tells a bit better what we are trying to say above.
