smirnovd's blog

Use standard track reconstruction with one layer of active Sti volumes excluded from tracking
Consider tracks with at least 40 TPC hits
In the skipped layer look for candidate hits within 5*sigma window in local Z and Y where sigma = track projection error (before the final track fit)

In TPC layers we see either zero or one candidate hit in thus defined window

There are some hits outside of 5*sigma window

The chi^2 of candidate hits is consistent with the cut applied in Sti standard track reconstruction
The measured efficiency can be described using the following components
P_measured = (P_instrumental + P_background) * P_corr * P_tracking

where

P_instrumental (or P_detector) is essentially the probability for the detector active media to produce a signal that can be recorded (Includes efficiency of hardware and/or low level thresholds, pedestals, ...)

P_background is the probability to have a hit within the window not produced by the considered track (Includes detector noise, other tracks, ...)

P_corr include cluster/hit reconstruction efficiency, uncorrected unknown distortions, inefficiency of known corrected distortions

P_tracking is the pure tracking efficiency, i.e. the probability to assign a hit to the track by the tracking algo.
Includes algo specific requirements such as number of time hits can be used/reused, hit acceptance/rejection cuts (chi^2), ...
Currently I measure (P_instrumental + P_background) * P_corr essentially assuming P_tracking ~ 1
P_tracking can be measured directly by matching track/hits WITH and WITHOUT the skipped layer

Extra support plots
How projected track position compares to the final track position with respect to the accepted hit?

How projection error compares to the final track error?
http://www.star.bnl.gov/~smirnovd/stihify/