This is still a work in progress, and more importantly, the pdf included here is not terribly useful without the comments that follow. 

I will add in track-based cuts shortly, but wanted to put this up as-is first.

The bottom line of all this is that work still needs to be done:  I will retool the preshower cuts, add in track-based cuts, and play around with the SMD cuts in greater detail in a coming post.

Update:  To be explicit, the following thresholds exist in the code:

 pam->SetMinClusterEnergy(0.3);
  pam->SetEtaRange(1.0,2.0);
  pam->SetMinPre1Energy(0.0005); //GeV.  changed to .0005 per Jan's request.
  pam->SetMinPre2Energy(0.0005);//GeV.  changed to .0005 per Jan's request.
  pam->SetMinPostEnergy(0.0005);//GeV.  changed to .0005 per Jan's request.
  pam->SetMinStripEnergy(0.0002);//GeV.

Cluster Energy is the minimum reportable.  If an cluster is created and has a summed energy below this threshold, it is not saved and not counted. 

Eta Range chooses the active region of the detector.  Clusters constructed outside of this range are not saved and not counted.

Pre1E, pre2E and PostE are not cuts in themselves - any response in those layers is counted, but the number of preshower or postshower towers considered lit is only incremented when the energy is above that threshold.  A preshower layer with less than .5MeV will count as .5MeV toward the preshower energy, but will count as 0 preshower layers for purposes of counting.

The min strip energy is the threshold below which we stop considering strips to be 'lit' in the same fashion, and also the point at which we will stop clustering on each side of a peak strip.


 

Comment Track (copied out of an email I originally planned to send to Jan):

Note that the windows are always selected so that there are NO overflow events.  Additionally, the two dimensional plots are always in the same positions - gamma in the upper left, muon in the upper right, pizero in the middle right.  They have labels, but these are frequently obscured by the histogram titles.

Cut profiles (one or two appear in this progression) count every cluster, regardless of which event it came from.  So I do not currently fill each H# that appears in your description.

page 1: this shows the cut profile for your suggested cuts (they're listed ont he right).   Note that the postshower cut is exceedingly damaging.
page 2: This is the cluster-sum postshower spectrum using the cuts before it in the cut list.  there's not a very strong signal for us to cut on, apparently.
page 3: Treating postshower energy as a fraction of shower energy, we can see a much friendlier cut -- postE<.001*towerE.
page 4: This is the energy-weighted center of reconstructed clusters with the first cut applied.  It's clear that the distribution of events is not uniform in X and Y.  This is perfectly expected, though:  we threw particles uniformly in eta-phi, and so we would expect more to have been thrown at x^2+y^2 close to zero, as is seen.
page 5: Here we see the center of reconstructed clusters (with first cut applied), but this time in eta-phi space (detector eta).  Note that it now appears roughly uniform, save for something close to a doubling at the highest eta - this is expected.  Since there are no higher eta rings, particles that would normally deposit most of their energy in the 13th ring instead leave only a smaller amount in the 12th.
page 6: To verify the explanation of the doubling above, here is the energy spectrum for potential gammas in the highest eta ring (or in that vicinity).  Note that it peaks at low energy.
page 7: And here is the same spectrum for the bin before it.  Note the distinct differences in the two distributions.    Most of the 'extra' events in the highest bin seem to be very low energy.
page 8: This is the high tower ET spectrum with the first cut applied.  We have a huge hadron peek at low ET, which is expected.
page 9: This is the one-dimensional energy-weighted eta of tower clusters.  We see the same eta dependence as on the two-dimensional plot earlier.  This has already been explained.
page 10: This is the one-dimensional energy-weighted phi of tower clusters.  It's a very noisy graph, but there doesn't seem to be any large-scale structure to it, which is good.
page 11: This is a close up of the previous page.  Here we can see the clear spike structure corresponding to the phi positions of single towers (which occurs when we are not close to the edge of a tower).
page 12: This is the X-Y position of clusters passing the first two cuts.
page 13: This is the eta-phi position of clusters passing the first two cuts.
page 14: This is the tower cluster ET spectrum after the first two cuts.
page 15: This is the tower cluster energy spectrum after the first two cuts.
page 16: Eta distribution of cluster centers after the first two cuts.
page 17: Phi distribution of same.
page 18: hightowerE/towerE after the first two cuts.  Since I only calculate eta after building the cluster, there is no difference between the eta of the cluster and the eta of the high tower, and so the relative energy is the same as the relative ET.  The high tower usually dominates the eighted cluster position, and in this case they are clearly equivalent.  When there is a high degree of energy sharing across a boundary, labeling the hightower with the eta of that tower's center isn't terribly physical anyway -- the real tower center should be right at the edge of the adjacent towers.
page 19: A two dimensional plot of cluster-summed postshower energy vs cluster-summed tower energy.  This lends support to the cut I suggested earlier (page 3 note).
page 20: Shifting to the cut I proposed above instead of the hard cut on postshower energy, we now have this cut profile.  We still have a factor of three reduction in the (hadronic) background, and now there is essentially no loss in signal (and pizero and electron backgrounds).
page 21: X-Y position of clusters that have passed the first three cuts (as described on the previous page).  It's fun to note that we can, from here, clearly see the three types of particles:  Heavily Populated=EM showers, Sparsely Populated=Hadronic showers, Unpopulated=MIPs.  (We expect the MIPs to be gone since they deposit the same energy in the postshower as they do in every other layer, and #layers is less than 1000.)  In future iterations, and with your permission, I won't bother to plot muons at all.
page 22: eta position of tower clusters.  not interesting save that it's a good check that we're not adding any funky eta dependence.
page 23: phi position of tower clusters.  Same note as previously.
page 24: cluster-summed pre1 energy spectrum after the first three cuts.  Note that the gamma signal peaks more sharply at zero than the EM backgrounds.  If we insist on any strict minimum here, we will necessarily chop out more signal than background.  The physical interpretation of this is that photons do not emit secondary radiation continuously, but discretely.  There is a finite chance for a photon to travel through a radiation length without emitting, and so there is a peak at zero.  For two photons, the chance of having /neither/ emit is smaller, and so we see a smaller peak.  For electrons, there is a very low chance - they're charged particles and emit continuously.  For these physical reasons I think a strict cut here is unwise.  That said, unlike the postshower, I don't have a quick suggestion for a replacement cut.
page 25: X-Y position of clusters surviving the first four cuts.  The labels start to get unreadable here, but the positions are the same as for all 2-D plots, and the titles I will describe here.  I intend to stick a TPaveText in one of the empty slots in order to list all the cuts more legibly.
page 26: eta of surviving clusters.
page 27: phi of surviving clusters.
page 28: cluster-summed tower ET of surviving clusters. Note that photons are now suppressed relative to the other EM showers for all but the highest ET.
page 29: cluster-summed energy in the second preshower layer for clusters surviving the first four cuts.
page 30: comparison of pre2 energy to tower energy (summed over cluster in each case)
page 31: pre2E/pre1E plotted vs tower energy (all summed, as always).  Note that the scales are different due to distant outliers in some plots.
page 32: X-Y position of clusters surviving the first five cuts (last cut applied is to allow only events with pre2E/pre1E>1.1).  Still no problems here.  For the hadronic showers it becomes difficult to assess troubles, since they're running out of systematics.
page 33: eta position of surviving clusters
page 34: phi position of surviving clusters
page 35: cluster-summed energy of surviving clusters.
page 36: pre1 energy of surviving clusters.
page 37: pre2 energy of surviving clusters.
page 38: We're starting the SMD profiling here (only on surviving clusters).  The SMD range is defined to be 40 strips in each direction of the energy-weighted cluster center (from the towers).  Here is the summed energy in the U plane.
page 39: summed V-plane energy.
page 40: sum of U and V energies.  Nothing surprising here.
page 41: V-plane energy vs U-plane energy.  Some closer inspection of the widths of these streaks shows maybe a little room for discrimination.
page 42: Summed SMD energy vs cluster energy.  We see no obvious difference between the three main survivors here - all have EM showers, so all look the same in gross shape.
(You also asked to see plots of the peak strip energy in the U and V plane - these are not present here (error filling the variable) but will be included on the next iteration.)
page 43: This plot shows the cluster energy of U or V strips in the first SMD Z position.  The peak at zero is a cluster that returns no -- or very little -- energy, implying there were no strips above the seed threshold in the entire area under the lit towers.  This may be related to a warning I was getting that the highest strip under a given tower range was showing up on the edge of the range of strips, where much of the shower energy would not be within the range.  It may also be a function of shower position, which I show in page 48. (And these two situations are likely strongly correlated) In any case, I'll look more into this.
It's important to note that I am not using the clustering algorithm that you suggested, but rather a loosened version of the one I got from Alan.  Here, I continue to add strips unless it falls below the minimum 'add' threshold -- 0.2MeV, or until the strip is more than 20% higher in energy than the previous strip.  This limit can be raised even higher if needed.

page 44: A close-up of the low energy region of the previous plot.  It's interesting to note that this region has roughly twice as many pizeros as photons.  This may very well be a feature of my clustering algorithm, which would tend to drastically lower the energy of pizero clusters at higher energy, when two peak strips are close to one-another.
page 45: The ratio of u cluster energy to u total energy.  As suggested before, we see that the pizero peak has a long tail toward lower fractions, suggesting that our cluster is stopping early.
page 46: A plot of U or V strip cluster energy in the second SMD Z position.
page 47: A plot of U or V strip cluster energy in the third SMD Z position.
page 48: A plot of geant_phi position (the actual thrown particle trajectory) for particles with less than 0.05GeV in the u or v planes. We see strong peaks along the sector boundaries, and noise elsewhere.  Some of the noise is a function of the cut I applied, which lets in hadronic showers which might validly have low cluster energies.  If I lower the cut any further, though, stasticis start to heavily limit it.
page 49: A better angle on the previous:  Now I'm cutting on the fraction of the u or v plane energy that is contained in the cluster (<10%).  For particles that span the boundary, I sum the plane energies over both sectors, so we would expect this fraction to be small there.  And yes, indeed we see a sharp peak at each sector boundary.  We may have to excise the sector boundaries
page 50: U-plane clsuter energy vs v-plane cluster energy, with  hightower ET>1, tower ET>4, postshower energy < .1% of tower energy, preshower >1.5MeV, preshower2 at least 10% higher than preshower1, and at least 1.5MeV each in the u and v planes.
page 51: summed u and v cluster energies vs tower energy with the above cuts applied.
page 52: distance between smd cluster X position and tower cluster X position, requiring the energies in the U and V clusters to both be higher than .4% (.02/5) of tower energy.  Note the single events way out on the x axis -- I don't know what causes this bad a mismatch.
page 53: distance between smd cluster Y position and tower cluster Y position with same cuts as above.  We dont' see the far-distant outliers in this plot, though there are still some events farther out on the tails than expected.  Primarily, the distance is +/- 5cm, which is reasonable considering that the towers usually don't give resolution better than the center-to-edge distance.
page 54: square of X distance summed withs quare of Y distance, requiring that the delta x be less than 20cm to suppress the most egregious outliers.  All the EM curves seem to have the same basic shape.
page 55: summed squared distance compared to the SMD cluster's eta position.  As expected, the agreement improves as we move to higher eta, where the tower resolution is better.
page 56: the summed squared distance compared to smd cluster phi.  To our great relief we see no dependence.
page 57:  The X-Y position of reconstructed clusters.  All cuts short of tracking cuts are now applied. 
page 58:  the eta position of reconstructed clusters.
page 59: phi position.
page 60: tower energy.
page 61: pre1 energy.
page 62: pre2 energy.
page 63: ucluster E vs vcluster E.
page 64: summed U and V cluster energy vs tower energy.
page 65: summed U and V clsuter energy vs preshower1 energy.
page 66: preshower1 energy vs tower energy.
page 67: fmax (hightower energy / tower energy) vs tower energy.
page 68: pre2E/pre1E vs tower E.