Est Performances using the AssociationMaker
These results were produced with the newly produced gstar files
with the correct number of wafers on the SVT layers. As far as I can
tell they look ok. SRS does not complain anymore about unknown wafer id.
1 - Low density results.
20 events of the low density file were processed to quickly check the
data quality. I used only one superpass requiring a hit in each SVT
layer for thes test. I asked also the z position of the main vertex to
be at less than 20 cm from the STAR center so basically I took all the
events in the file.
Maria notices a strange distribution of the
MC tracks in phi attributed to the wrong SVT geometry in gstar.
Asking for at least one hit in the MC track gives the following distribution :
Asking for three hits in the MC track gives the following distribution :
So my guess is that we can consider this problem as fixed.
Now the Efficiencies. Maria defines two efficiency histograms in her Maker. One is :
mBestMatchMcTrackPt
mMatchedTracksEfficiencyRcMap = -------------------
mRcPrimaryTrackPt
and the other is :
mBestMatchRcTrackPt
mMatchedTracksEfficiencyMcMap = -------------------
mMcTrackTpcSvtPt
The following histograms are for respectively at least one hit and three hits in the MC tracks (one superpass implies 3 hits in the RC tracks).
The RcEfficiency is the red curve whereas the McEfficiency is the black curve.
Using the Est own evaluator I got an overall efficiency above 90% which is consistent with the McEfficiency
2 - Standard density results.
I
have processed 5 standard density events in the same manner as the low
density file. Only one superpass and 3 hit requirement in the association
maker.
From the Est evaluator, I obtain these results cumulated over the 5 events :
StInfo: ******************** EST Evaluation Summary ************************
StInfo: Total number of events : 5
StInfo: Total number of ideal tracks 3403 2630 773 (all/prim/seco)
StInfo: Number of ideal tracks/event 680.6 526 154.6 (all/prim/seco)
StInfo: Total number of good tracks 2396 1972 424 (all/prim/seco)
StInfo: Number of good tracks/event 479.2 394.4 84.8 (all/prim/seco)
StInfo: Total number of bad tracks 291 174 117 (all/prim/seco)
StInfo: Number of bad tracks/event 58.2 34.8 23.4 (all/prim/seco)
StInfo: Tracking efficiency (%) 70.4085 74.981 54.8512 (all/prim/seco)
StInfo: Purity (%) 89.1701 91.8919 78.3734 (all/prim/seco)
QAInfo:est : Real Time = 144.30 seconds Cpu Time = 143.88 seconds, Entries = 12
Which is significantly lower than what I used to have when looking at
year_2a fixed vertex events.
Typically I obtain efficiencies of
83% / 92% / 50% for all / prim / seco tracks. The new file is a
mixture of -15,-5,+5 and +15cm vertex displacement in z so it should
affect the averaged efficiency. I looked at the Est histograms where
the overall efficiency is calculated versus the vertex position in z. The
histograms are in this ps file.
There is a z position dependance of the efficiency but
which does not explain the differnce with the year_2a file.
I have saved into this ps file all the histo. maria has
defined in her version of the AssociationMaker. I have enlarged the
eta cut to +/-2 and lower the minimum number of tpc hits to 5 (as in est evaluator).
The results are close to the results from the Est evaluator but they
are not consistent. For example, the number of ideal tracks is not equal to the
number of MC tracks.
Last modified: Wed Mar 28 14:45:37 CEST 2001