So, after the extensive 3/4 trigger efficiency study, we need to figure out just how much of a contribution it makes to the overall HMS trigger efficiency.

What we need to do is look at the efficiency of all the legs that make up the electron trigger and see how bad the worst case is (between elhi and ello). Below is a schematic of the trigger (we're interested in the El Real part) that I stole from John Arrington's thesis, but I'm not smart enough to crop.
Using the elastic runs (51861-51887), I figured out the efficiencies for all the legs. The 3/4 and the STOF efficiencies are not on this plot, but they exist on the SCIN and STOF pages.

To calculate the individual efficiencies, I got a clean electron sample and looked to see what fraction of that sample fired a given trigger.
The cuts applied to *all* the samples:
abs(hsdelta)<12
abs(hsxptar)<0.08
abs(hsyptar)<0.04
abs(w-0.938)<0.02
Next, here are the cuts to get an electron sample for each trigger.

Trigger ------	Cuts
CER	ELHI.and.(hsshsum>0.7)
PRLO	STOF.and.SCIN.and.(hcer_npe_sum)>5)
PRHI	STOF.and.SCIN.and.ELLO
SHLO	STOF.and.SCIN.and.(hcer_npe_sum)>5)

I might have to redo PRHI, because while we require the ELLO trigger to fire without requiring its calorimeter leg (PRLO), we're not strictly speaking getting electrons only, since we only require CER to *fire* without specifying the number of photo-electrons. It's already pretty high, so it can only get higher. (UPDATE: it went up about 0.2%). Update 2: These are the latest plots. The previous generation (without the w cut, and a cut on hsshtrk instead of hsshsum can be found here. The cut on the elastic peak made a noticeable difference in some of the efficiencies.

Here's the same plot with momentum on the x-axis.

Here's a similar plot, but with regular data where the pi/e is not too bad. Also, I tightened up the hcer_npe and hsshtrk cuts. This one did not change much from the previous incarnation, but I didn't really ask it to change much.

Hardware/Software Efficiency Check

Once we're analyzing data, we make software cuts (hcer_npe>some number and hsshtrk> some number). With this next efficiency check, we want to make sure that all the events that pass a given software cut also fired that given trigger (hcer_npe>2 must have fired the Cherenkov, right?)

Here are the results

To see this with the HMS momentum on the x-axis, go here.

To calculate the efficiency of the ELLO trigger is non-trivial since the STOF and SCIN legs are correlated: SCIN is a subset of STOF. We know that if SCIN fired then STOF fired for sure. But what about the other way around?
The value that are < 1, are not qutie elastic runs. They're part of the "scan."

So, with our software cuts, the Cherenkov will always fire, PRLO efficiency is ~.9997, and the STOF efficiency is ~.9985. The overall ELLO efficiency is then approximatedly .9997*.9985=.9982 The real efficiency is probably ELLO=SCIN*STOF*PRLO+SCIN*STOF*(1-PRLO)+SCIN*PRLO*(1-STOF)+STOF*PRLO*(1-SCIN). For a more detailed discussion, go here .

So, it was pointed out to me at the meeting today (5/18) that I shouldn't use a w-cut on these runs to get my stof/scin efficiency, since that will just pick events that almost certainly fire both, so yeah, the efficiency is 1.

Redoing that exercise without the w-cut, but still using the elastic runs and a production run at 2.67Gev, 32 degrees, I get the following (definitely not 1):