Short Biased Minutes of

W meeting 13/10/1999 in 40/SS-C01 from 10 am to 2(!) pm

1) Presentations for Crete : introduction + the mass (Ann++)

We are contributing to the Crete workshop in several ways both by giving a talk over there (12 over a total of about 35) and by contributing to other's presentations. As the idea is to prepare the best possible LEP averages which means a) understand each other errors b) understand which method gives the best error, we decided to give our share of understading in the informal discussions. In other words we will try to explain to our DLO colleagues what we have understood and believe it is solid. This implies an effort to publish rapidly our W mass after Crete (and the minutes writer is a strong supporter of this last point!). In summary this are the new developments we will try to explain :

a) Jeremy's ISR treatment (ALEPH note in preparation)

b) Semileptonic fitting (1D vs 2D vs 3D see Oliver later)

c) Fragmentation "don't use Monte Carlo only" (see Jason later)

d) Reducing BE correlation effect on data (see Jorgen recent Aleph Note)

e) BE a la Delphi (mixed events) with Aleph data (Franck)

We will also informally mention which are our present expected error for some systematics relevant for the discussion (fragmentation, BE,CR,..)

It turned out during the meeting that some plots from c) and e) were not yet ready so we decided to have a special continuation of this meeting on Monday, the 18th, afternoon. During this appendix we will see also some transparencies Tim and Helenka are preparing for their TGC and tau talks in Crete.

2) Fragmentation (Jason)

Good news : a bug was found in the way we were treating radiative semileptonics in Herwig, which caused non-conservation of energy. From preliminary studies (Oliver) there are strong indications that this is the cause of the doubling of the fragmentation error when comparing Herwig-Jetset at 189 GeV with respect to 183 GeV. As far as Crete is concerned Jason showed a detailed workplan from the 4 exp mainly on MC comparisons, which however could not quite fit into he Crete timescale. Concerning our MC/data comparison method as mentioned above not all the plots were ready for the semilept so we will continue on monday. A discussion followed about what's worth showing and the decision was to go along with showing what we aim to do about data/mc comparison. This decision was coupled with the intention to publish asap after Crete .....

3) FSI (Frederic)

Frederic showed the agenda of this part in Crete. For the Aleph talks (Franck and Jorgen) see above. There will be more talks focusing on the implementation of different models in the MC, Aleph technique on Delphi data, string effect method by L3, etc.

4) Online Cross Section (Anne)

We do have a cross section also at 202 GeV from the collected 12 pb-1 ! See the web page to admire the new plots extended after 200 . We are well in shape to provide plots LEPC ready ! We are happy with the current status of MC production and with the decisions about the 200GeV class 16 bug .

5) Some observations about BE (Jorgen)

Jorgen showed variations of pion multiplicities, pairing efficiency, clustering eff. with/without BE correlations. However after the meeting it was found that our Luboei samples are affected by "non-random number conservation" when changing options about BE (i.e. also fragmentation developed differently on an event-by-event basis). This will be fixed thanks to Brigitte. Therefore this study of Jorgen proved useful to find this "feature" but now it has to be repeated.

6) 4q FSI systematics (Evelyn)

Evelyn reviewed the work she has done for the systemtic errors on the 4q mass with the standard MC method. The CR error is dominated by far by the SK1 sample with the arbitrarily 30% reconnected events (and Oliver correctly pointed out that this should be 27% at 189 GeV since it is energy dependent) and it is now 62 MeV. When the statistically limited evaluation from BE is added we get 65 MeV from FSI. The feeling is that if we go along with this method we should provide also the plot with the depedence vs reconnection probability.

A new study of Eric shows that this dependence is not affected by the preselection, so it can be safely used. Eric also tried to find a "reconnectometer" to see if some extreme reconnection probability can be ruled out by data. Several variables were used and some (minum jet mass for instance) look very promising.

The new event selection by Evelyn (purer preselection, then simple weights using 4 variables) is final and working. A note will be written and the code made available. An efficiency of 85% with 78% purity can be obtained (similar to Opal, it can be compared to 92%-79% with Aleph NN14).

Evelyn started to look into the L3 particle flow method but got into troubles with the azimuthalm angle. To be clarified with L3 in Crete ....

7) 1D-2D-3D reweighting in the semileptonic channel (Oliver)

Oliver has added to the 2D reweighting (the two masses which were used are the 2C fit mass and the 1C mass on the hadronic side) the error slicing which was shown to be promising by Helenka in Siena. He has started first with 2C mass plus slicing to test with a 2D technique the effect of event by event error. With 189 GeV data this works, contrary to what happened at 183 GeV (see Jason talk in Annecy). The reason for this different behaviour (183 vs 189) is not yet understood. At this point Oliver went to 3D reweighting getting the best error. A nice plot with the history of the improvements in error versus the various methods was shown. We got 15% improvement in the stat error since Moriond ! The idea is to show this (for MC only) in Crete .

8) lvlv W mass (Djamel)

Djamel showed his preliminary results at 189 GeV on the lvlv channel using the method pioneered by Severine and Pierre. He has studied the calibration curves and realized he needs higher statistics for the reference sample. The work is in progress.

9) Semianalitical study of the W mass statistical error (Andrea)

Welcome back Valassone ! The audience was a bit tired at 1.30 pm but still alive to listen to the semianalytical study of Andrea. For details read the transparencies which are quite clear. The basic idea is that the statistical error depends only on the functional form. Andrea has demonstrated (in practice with infinite statistics) that in the 4q channel

a) it is not worthwhile increasing the window with respect the one we are presently using

b) the best error we get with NN14 cut at 0.3, due to the interpaly of eff/purity, confirming the Monte Carlo studies done with reweighting 2D

He is ready to use this interesting new tool to answer to further questions, stay tuned !