The informal meeting started with Claus Grupen, who showed his catoons which summarize exactly (and graphically) the status of Higgs searches in ALEPH and LEP. Systematic checks in HZ->tautau (Jason Nielsen) ----------------------------------------------- Jason sumarized the systematic checks have been done in HZ->tautau channel. Studies in hA->tautau will be followed up later. For 144pb**-1 of data taken this year, 8.9 events were expected ( 0.5 eeZ + 0.7 qq + 4.2 ww + 3.6 ZZ ) and 9 events were observed. However for reconstructed higgs mass greater than 110 GeV, 3 candidate events were oberved with 0.5 events expected ( 0.01 eeZ + 0.07 qq + 0.3 WW + 0.1 ZZ ). The event displays of these three candidate events and some details of event properties were shown. A Sical/Lcal object around 20 GeV is seen in (Run 56620 Evt 1910 MH = 117.2 ). Some concerns are raised on if ISR effects are correctly modeled in our MC. Further checks are suggested, for example by using data taken in previous years. On other properties like fit estimator and isolation angle got from 9 observed candidates, general agreements between data and MC are obtained within very limited statistics. Loosen cuts to preselection level to check those agreements were suggested. 4jet NN status ( Jinwei Wu) --------------------------- Follow up the studies shown in previous meetings, Jinwei showed a set of recent systematic checks related to the 4jet excess. Currently 30 candidates are observed with 31.7 expected. 7 candidates observed ( 3.7 expected ) at reconstructed Higgs mass above 110 GeV are mainly responsible for the 3.3 sigma excess in SM higgs search. At preselection level, event quantities measured from data are compared with MC prediction. Among them, y34, m12+m34-91.2, (1-NN3)(1-NN4) (and many others ) showed general agreements. No threshold excess is seen in m12+m34-91.2 distribution at preselection level, which may ease the suspicion ( a little bit ? ) about the excess observed in 4jet is due to low mass candidate events been artificially ( nobody knows how ) pushed to threshold. -- Actually previous data doesn't support such "threshold effect" either. A slight shift towards b-like range in data is observed in the sum of the 4 btag NN outputs. Jinwei tried to use last year's (d0,z0) smearing value, end up with almost no effect and the shift still persists. Check the effect from gluon splitting is suggested on this matter. Jinwei showed an interesting check: Rejecting candidates with a pairing in 10 GeV ZZ or WW ellipse. Above 110 GeV reconstructed mass, the cuts reject 37% of the background and 42% of the signal, observed events are reduced from 7 to 3. If assume all candidates are from background, probability 3 or fewer remain is 23%, and if assume all candidates are from signal, probability 3 or fewer remain is 33%. So the test (which seems good) behaves counter-intuitively and has almost no discriminating power. The next test is to see the difference between masses from 4c fit vs 5c fit. The mass difference of the high mass candidates are all within 1GeV. In addition, the 5c fit probability is good ( almost flatly distributed between 1-100% ). A "sideband check" is also performed. Taking events just below the NN cut ( 0.4-0.763 ), a clear excess for reconstructed higgs mass above 110 GeV is seen. This may indicate, above all, the excess observed may not be a higgs signal. Higgs candidates in the eyes of a WWer (Oliver Buchmuller) ---------------------------------------------------------- Oliver showed his critical look of the high mass candidates from pure kinematic point of view. To sumarize: (event by event information can be found from his transparencies) for the 8 candidates in total, 6 events with very poor chisqure for HZ hypothesis. 2 candidates have nice chisqure of fit, however the best fit is equal mass hypothesis (m1=m2=102.1 and 100.3 GeV/c^2 ). At this level, the results are bit surprising and somewhat contradict with Jinwei's result. To have a complete picture, following checks are suggested: i) repeat the fit with HZ signal MC ii) Oliver used P scheme for jet clustering which is a bit different from E scheme currently used for Higgs analysis. so it is worthwhile to repeat the fit with E scheme. iii) repeat the fit with ZZ and WW MC. Pairing studies with a neural network (Yuanning Gao) ---------------------------------------------------- Yuanning showed a very preliminary pairing study for 4jet, using NN technique. For a four-jet final state, there are six possible combinations to form an HZ event. Frequently one may see one combination can be consistent with ZZ or WW even from signal MC event (ref: Jinwei's talk). Yuanning pointed out to have a complete picture on how a candidate event is signal or background like, a multi-dimensional PDF which takes into account most important information for both signal and background processes is needed. In practice, such a PDF can be contructed by the aid of NN technique. A 12 variable NN ( 6 current NN outputs corresponding 6 possible combinations and 6 Z-masses with each combination. The combinations are ranked by the values of their corresponding NN output ) is trained at 206 GeV, with 114 GeV signal against QQ/WW/ZZ. At preselection level, the NN output generally shows agreement between data/MC. An excess at signal range is also seen from this NN (at NN output above 0.8, 11 events observed with 6.5 expected ). The current selected high mass candidates are also quite signal-like with this NN. More work is clearly needed to explore the potential of the study.