Test 4: visual comparison with 2nubb data (plus efficiencies)

[Shoram444]

Overview

Simulated data-set of 10k events of 2nubb, reconstructed with both algos.

Data-cut efficiencies:

data-cuts	CAT [%]	TIT [%]
2 tracks	30.75	33.88
+ 2 foil vtx	25.39	25.34
+ 2 calohits	15.05	14.72
+ 2 associated calohits	13.71	11.26
+ vtx <50mm	9.92	8.48

Gallery:

Both algos work fine

CAT	TIT

In these "standard" cases a quantitative analysis will have to be made to determine which has better precision.

Only CAT successful

CAT	TIT

In general there seem to be 2 ways this situation happens:

1. Sometimes TrackIT algorithm "cuts" the line too soon. In the first row in the table above it can be seen that the reconstructed track from the TrackIT algorithm suddenly ends in the 3rd row of tracker cells. In the falaise vertex extrapolation, however, there is a condition that it will attempt to extrapolate vertices only if they are on row 1 or 2 of tracker cells.
2. Sometimes multiple tracks are detected by TIT where there should be one. This is a common problem for both algos, though.
3. Sometimes unlucky kinks happen...

Only TIT successful

CAT	TIT

Comments:

1. For more complicated events - very long tracks with kinks and multiple scatters, it seems TrackIT is having better performance.
2. Kink reconstruction works for TIT, whereas it's completely lacking in CAT. But still a lot of work is needed!