- Efficiency table for backgrounds ANSWERClose
: included in v6
- There is some concern over the problem of the gain switch and ECAL
overcleaning in MET. ANSWERClose
We actually had the same concern and for that
reason broke up the cumulative plot into several run ranges, seen in
Figs.17 and 18 up to v4 of AN-10-314. The most recent run
does not have this MET overcleaning problem. For the middle term, we
use the re-reco, which just became available.
- Is the dominant background at high MT mostly off-shell W/W* or is it
boosted W due to hard jets. ANSWERClose
Looking with fireworks most events look pretty clean. To be
more quantitative, we will plot the number of jets and the jet pT
vs. MT (plots will be available soon).
- Does our background techniques predict any other cut inversion to
give us some estimate that could lend credence to our techniques? ANSWERClose
This is an interesting question. We want to apply our present
technique to the ET/MET region outside of our signal region
(ET/MET>1.5) which should be QCD dominated and compare the template we
get from data to MC.
- Is there correlation between MT and electron isolation that may bias
the shape? ANSWERClose
Unfortunately, we do not have a smooth isolation variable
to see how it varies with MT (due to HEEP selection). We can select
some variables (had2 and track isolation) and check the trend of other
two isoaltion variables (had1+EM vs MT) vs MT.
- The final MT plot should contain all systematic uncertaintiesANSWERClose
Right now only the error on the electron scale is included in
Fig.16. We are working on a (readable version) of the final MT plot
with an error band incl. all systematic uncertainties.
- Can we exploit MT shape for cheap gains in expected limit by
changing the nominal window? ANSWERClose
No, as we optimize to achieve the best expect limits.
- The "Depth" descriptions are confusing. Depth2 is really
something different depending on whether we are discussing Depth1 or
Depth2 isolation? ANSWERClose
it is indeed confusing It mainly concerns the regions,
towers 1-17 are entirely covered by depth1 and only towers 18-29 may
be split in two depths since more energetic particles are expected in
this region
Over the weekend a new version of the PAS and the AN were provided, including the comments and suggestions previously discussed (see firstset of answers).
Here we address the new comments by Jeff and some remaining ones which could not yet been covered in the first set of answers since we were still running over the data.
We also attach some slides with plots for the following questions.
1) Are the outlying events with large MET differences correlated with large MT, or large electron ET?
When pfMET disagrees with tcMET or caloMET, is it when the PFelectron and GSF electron are very different 4 vectors?
I worry that there could be an inconsistency between the electron used by pfmet and the gsf one which can induce an artificial met tail.
In the inclusive W analysis this was ignored because it was uncommon and did not affect the core resolution for on-shell W's;
but in your case it could significantly spoil the mc description of the W pfMT tail.
ANSWERClose
Although we don't see major disagreements between data/MC in the MT tail, data events were checked regarding large differences in the MET values from different reconstruction algorithms.
Some plots concerning differences between pfMET and tcMET are shown on pages >=10 (after full selections). Both algorithms usually agree, on seen on pages 11,13.
We found a handful of events with pfMET- tcMET > 25 GeV. Those were checked with fireworks, the event displays are in the attached slides.
2) I am having a hard time reconciling the content of Table 5 with Table 9.
2218 multijets survive Table 5 cuts, but then Table 9 starts with4687 events with MT > 50?
Can you also introduce a column to Table 9 with no MT cut (should that correspond to the rightmost column of table 5?)?
ANSWERClose
For this, we lowered the MT cut on Table 9 to 45 GeV (that's the minimum given our requirement of ET_electron>30 GeV and the dedicated kinematic cuts which effectively place a threshold on MET).
Note however, that the main difference between Tab.5 and Tab.9 are themethods to determine the #evts.
While Tab.5 is MC out-of-the-box,Tab.9 contains the bkgr from data-driven techniquies. Indeed, we see roughly x2 more QCD in data than in MC.
3) The W background is normalized to data, but the shape is from pythia6 MC.
Are the shortcomings of Pythia vs. NLO generators in describing the MT tail understood?
ANSWERClose
A comparison of Powheg and Pythia (without pile-up) generator for W was shown during pre-approval.
We included the plot again in the attached file on page 9. Both samples agree reasonably well, with low stats for high MT.
4) Can we exploit MT shape for cheap gains in expected limit by changing the nominal window?
ANSWER: no, as we optimize to achieve the bestexpect limits.
Perhaps you didn't understand the question? I believe you are close to optimal with a one-bin cut-and-count design;
the question was whether a multi-bin design, which exploits the fact that ~1.3 TeV W' has MT peaking far above your minimum MT, gives you extra sensitivity
(since for example the 0.5 background events predicted are concentrated at the low end of your search window).
A simple way to satisfy this request is just to tell me how much better the expected limit would be if there was 0 expected background above, say, 500 GeV? 450?
This sets the ceiling for how better you could possibly do.
ANSWERClose
OK, to answer your question: with the current setup, we have expected limit = 1.31544 (observed = 1.3583).
When setting number of background events to 1e-8 (about zero), we get an expected limit of 1.37994 for window MT > 450 GeV and 1.37519 for MT > 500 GeV.
So, we would gain in expected limit, but not much in observed limit since we see zero events.
Does this answer your question?
5) Earlier comments, had asked for the number of jets in high MT range.ANSWERClose
We attach some distributions on the number of jets after full event selection.
Page 3 demonstrates that the majority of events has a small number of jets and they mostly cluster in regions of low MT.
A comparison of the #jets vs MT (page 4) for data and the W-MC shows good agreement.
6) Earlier comments has requested to study the high ET tailANSWERClose
We provide plots of E/p for different ET bins on pages 6,7 (after full selection), comparing data and QCD from MC, demonstrating good agreement within statistical limitations for EE and EB
7) For the ttbar section, note that the results are still for 22/pb and to be updated this week for the full 34.9/pb statistics.
Would it be sufficient to merely normalize the ttbar background to the single tag events with MT>100?
And assign a generous systematic for b-tag efficiency?
ANSWERClose
Following Jeff's suggestion, we tried to use a mixed data-MC approach, taking n1 from data and the b-tagging efficiency from MC and assigning a generous uncertainty on the b-tagging efficiency.
With this approach, we obtain an estimate of the number of ttbar events for MT > 100 GeV of N(ttbar) = 21.8 +/- 19.1 (in agreement within errors with the MC expectation), assuming an uncertainty of 50% on the b-tagging efficiency.