CAVEAT: The analyses described here have to be considered just as "exploratory" analyses, not much more reliable than parton-level-analyses with some detector smearing.
FAMOS and ORCA single top analyses are described here.

CMSJET fast simulation analysis:

~/cmsjet/cmsjet_pyt_m.f reads the filename.ntpl ntuples, giving as output the column-wise ntuples filename.hbook containing the following informations:

TRANCH   * CIRC
TRANCH   * TREN
TRANCH   * TRMA
PTMIS    * PTMT(2)
PTMIS    * PTM(2)
PTMIS    * PZNU(2)
LEPTONS  * NLE_CWN
LEPTONS  * PL_CWN(4,NLE_CWN)
LEPTONS  * KL_CWN(NLE_CWN)
LEPTONS  * MRL_CWN(NLE_CWN)
LEPTONS  * LISOL_CWN(NLE_CWN)
GAMMAS   * NGAM_CWN
GAMMAS   * PG_CWN(4,NGAM_CWN)
GAMMAS   * LISGAM_CWN(NGAM_CWN)
JETS     * NJG_CWN
JETS     * PJG_CWN(4,NJG_CWN)
JETS     * MRKJET_NJG(NJG_CWN)
JETS     * BTAG_NJG(NJG_CWN)

The useful ones are:

PTM(2): Et missing along x (1) and along y (2)
PTMT(2): "true" Et missing (i.e. due to neutrinos)
PZNU(2): the two solutions for the P along z of the neutrino, in the M_W constraint
NLE_CWN: number of tagged leptons (e,mu, not tau)
PL_CWN(4,NLE_CWN): lepton 4-momentum
KL_CWN(NLE_CWN): lepton "truth" (+1: e^-, -1: e⁺, +2: m^-, -2: m⁺)
MRL_CWN(NLE_CWN): PYTHIA code of the lepton "mother" (W^+-: +-24)
LISOL_CWN(NLE_CWN): lepton isolation flag (0: isolated, 1: not isolated in the tracker, 2: not isolated in the calo, 3: isolated neither in tracker nor in calo, 4: isolated in a jet)
NJG_CWN: number of jets
PJG_CWN(4,NJG_CWN): jet 4-momentum
MRKJET_NJG(NJG_CWN): PYTHIA code of the parton originating the jet (b: 5)
BTAG_NJG(NJG_CWN): b-tag variable of the jet

All the analysis parameters and switches for ~/cmsjet/cmsjet_pyt_m.f, as well as the input and output ntuple files, are given by cmsjet_pyt_m.dat (also needed is myrandom.dat, which has to contain a 10-digits random seed).

I use a slightly modified version with whom I also save number and 4-momentum of the true top quarks. To compile (on lxplus) use this Makefile.

PAW macro ~/scratch0/top/top.kumac opens the ntuples from CMSJET and plots the required variable comparing different event classes.
The analysis file ~/scratch0/top/readcwn.f (to be compiled with ~/scratch0/top/readcwn.csh), whose parameters and common blocks are in ~/scratch0/top/readcwn.inc, computes how many leptons (from W->l or not) are isolated, how many jets (from b quark or not) are b-tagged, and how many events are selected for different event classes.
Currently the following preselection criteria are chosen:

1 isolated lepton (Pt>20 GeV)
2 jets with Pt > 30 Gev
No other jet above 20 GeV (this is the detectability threshold, but CMSJET reconstructs also very low Pt jets)
at least 1 b-jet (b-tag variable > 1.0; in CMSJET it is the second highest i.p. significance in the jet)

The PAW macro ~/scratch0/top/top2.kumac permits, after the execution of readcwn, to plot some variables after preselection.
To use Mreco (invariant mass of b+l+"nu") one has to choose one b-jet and one solution for the neutrino (since the W-mass constrain gives a second order equation). For the neutrino I choose the solution that gives the smaller |Mreco-Mtop|. For the b, I choose the "best" b-jet (i.e. the one with the largest b-tagging variable).
This gives an ambiguity for s-channel signal and for ttbar and Wbbbar backgrounds. This is not a serious problem when looking for single top vs. ttbar and Wbbbar, since s-channel is a small signal component (and the ambiguity in the choice broadens the Mreco distribution for the backgrounds, making them easier to throw out), but a better criterion should be found when trying to discriminate among the signal components.

For single top selection I cut on Mreco and Ht. Using the set of cuts in file cuts_singletop.inc (selectionmode=1 in ~/scratch0/top/readcwn.inc) I obtain the following results: sn_singletop.out (table preselected, table selected).

If I restrict the search to the s-channel (better theoretical errors, and bigger improvement on the main systematics is foreseen from high statistics LHC background studies, e.g. qq'->W) I impose two b-jets in the final state and use also the eta and momentum of the less b-like jet, and its angular distance from the most b-like. Using cuts_schannel.inc (selectionmode=2 in ~/scratch0/top/readcwn.inc) I obtain: sn_schannel.out (table preselected (2 b-jets instead of 1), table selected).

It's quite easy to select an almost pure t-channel sample (its advantages -and drawbacks- for studying single top polarization are discussed here): I impose only one b-jets in the final state, anti-tagging the second jet, and reject all events in which the angular distance between jets is below 3.2 (very few non-t-channel events fulfill this requirement). Using cuts_tchannel.inc (selectionmode=3 in ~/scratch0/top/readcwn.inc) I obtain: sn_tchannel.out (table selected).

(Results obtained with sn.f, which contains the theoretical cross sections and reads the efficiencies from readcwn, writing the results as a LaTeX table. The script sn.csh executes readcwn for all the signal and background channels, and at last launches sn.)

Trigger studies:

We want to see the effect, on the selection efficiencies and on the shape of the variables, of the cuts listed in the DAQ TDR at page 285.
In trigger.f I gather several selection streams.

Last developments (after CPT week, may'04):

For s-channel the criterion of the "best" b-jet for top reconstruction is ambiguous, since both jets are real b-jets. This is not a problem for single top search, but it is for s-channel search.
This figure shows that, while for t-channel the goodness of the choice is clear, for s-channel the top mass reconstructed with the worst b-jet is almost as precise as the other.
I decided to exploit this difference, with a cut on Mreco("wrong")-Mreco("good")<50 GeV. This improves S/B from 0.21 to 0.24, and s/t from 0.39 to 0.47.
Polarization (in the top production) in t-channel: I boost the lepton in the top rest frame (caveat: for s-channel there is still the ambiguity pointed before!) and calculate the cosine of the angle between the lepton and the recoiling jet (i.e. the least b-tagged jet). This makes sense for t-channel only, so it's useful to select a pure t-channel sample for this kind of studies.
Lepton distribution in t-channel clearly follows an A+B*cos(theta) law, showing that polarization is present (as predicted in particular by electroweak theory, in which g_L=1 and g_R=0), while the backgrounds do not (see figure; the plots are: s-ch., t-ch., Wt, tt, Wbb, weighted sum (taking into account the expected cross sections and BR(e,mu)); the last bins are not included in the fit since they are depleted by the lepton isolation cut). The normalized slope (i.e. B/A) for t-channel (from the fit results) is 0.87+-0.04, in rough accordance with the theoretical predictions (i.e. ~0.9). After t-channel selection: there seems to be a slight bias towards positive slope (probably due to the cut on the angular distance between jets) even for the non-polarized backgrounds (see figure); for t-channel the result is B/A=0.95+-0.08, and for the weighted sum it is still B/A=0.95+-0.08, so the contamination appears to be well under control.
A dedicated PYTHIA t-channel sample shows a flat angular distribution; since PYTHIA doesn't take into account the polarization of the produced quarks, this is a proof that the unflat distribution for the TOPREX sample is not an analysis artifact.
More details, further discussion and theoretical motivation here.
I generated 10000 Wcc->lncc events, using a modified code to force cc instead of bb. CAVEAT: use CMKIN_3_1_0 and not CMKIN_2_1_0 to exploit this trick.
Applying CMSJET to this sample, the results suggest that this background is well under control, as can be seen from the output files from sn.f linked above.
I generared 20000 tt->2l2nu2b events. This is a very dangerous background, since when one lepton is outside acceptance the rest of the event closely resembles the signal. I have to introduce an upper cut on the missing Et.
Other backgrounds (20000 events each): WW (inclusive), WZ (inclusive), ZZ (inclusive), WZ->lnbb. The last one suggests a cut M(j1j2)>110 to reduce the contribution from Z->bb, which is of course peaked around the Z mass. This is not a very consistent background, but this cut is almost for free (very few signal events fall below this limit).
The parameter mrecoswitch3, in ~/scratch0/top/readcwn.inc, if set to 1 tells the program to choose the reconstructed mass closest to 175 GeV among the four combinations of neutrino z-momentum and b-jet. This significantly improves the resolution for Mreco and Preco, but S/B decreases since also background gathers around 175 GeV.