We can use a high-pt central muon sample to try to address both stub-finding efficiency and central tracking efficiency.
Assume that the muon represents one leg of a Z and calculate where the other leg would go (2 cases) if the Z were at rest. Look in a window about that for a track of about the right momentum and then look for a matching IMU stub. This procedure has been used before to estimate our stub-finding efficiency. No problem here.
Assume that the muon represents one leg of a Z and calculate where the other leg would go (2 cases) if the Z were at rest. Look in a window about that for an IMU stub. You expect a lot of fakes, as we see below. Now look for a track. If none exists (most of the time), look for hits that almost form a track. This requires a lot more knowledge about trackfinding.
I generated realistic Z's (width, the CDF 4561 Pt distribution) and used the Pt=0 estimate to predict the other leg position. If I required that the central (η nearer to central) leg have a Pt > 15, I found distributions for deviation in η and φ.
If I demand that the deviations from nominal in φ and in η are both less than 1.5, .857 of the original (Pt > 15) tracks survive. This is an extremely wide φ window; it represents the entire opposite side.
So we can try to tighten things up, with accompanying loss of efficiency. With loose ( < 2) cuts on the other variable, you can see the relative efficiencies as you cut harder in φ or η. (Note that what is plotted is not absolute rates, but rates relative to the specified cuts.) Pt > 15, of course. If we cut with deviation in η less than 2 and deviation in φ less than .5, our efficiency is .764, which isn't that huge a loss but helps clear the clutter away. A cut of .3 gives .665. Using some old plots I estimate the current rate of IMU stubs as being about 2 per event, which would put about .07 into a (2*) .3 window per event.
Modified 14-October-2005 at 15:37
http://hep.physics.wisc.edu/~jnb/imu/17Oct2005