[Brahms-dev-l] Re: FS tracking efficiency with the reference method.

Hi Jens Ivar,

Yes it is a bit strange/unphysical with the discontinuities between 100% 
and 80%. I believe that Pawel has once showed that the efficiencies works 
in the sense that FFS and FS is similar with efficiencies. I am also 
convinced that it works since Djamel and I have compared FS to FFS and 
different FS/FFS settings to each other where the efficiencies were quite 
different and found good agreement.

So likely for some reason we don't quite understand the tracking "works
better" in some part of the detector than others or the statistics is
low or .... It is at least also somehow clear that the 3d histogram is too 
big for the low number of tracks that makes it all the way through the 
FS, and maybe you could reparamatrize this, summing over the top 0-20% 
and summing over fx the coordinate in a detector to get an efficiency 
that you can apply track by track. 

What Djamel and I did was to calculate the average tracking efficiency as
a function of momentum (a profile of the single track efficiencies) and to
use that for the correction (actually using the average where the counting
statistics was low) - this is at least be continous. I think we got the
same results no matter what we did, but it makes it a lot easier to
understand what you do.

Cheers,
   Peter

On Thu, 29 Jul 2004, Jens Ivar Jordre wrote:

> Hello out there.
> 
> Here comes a long mail from the northern part of the collaboration. I'm
> fiddling with the FS tracking efficiency calculated for the 2001 AuAu in
> hope that this is the end of the analysis for my thesis. Those
> interested better get a big cup of coffee before proceeding. :)
> 
> I still have problems figuring out how to interpret the efficiency as
> calculated in the reference track method output file. Mine is called
> "Effic5361_5972.root" which I assume other people are using too.
> 
> First, let me describe how I extract _each_detector's_ efficiency from
> the file. It is close to identical to the way done in the
> EfficiencyCalculator and the BdstFsAna classes for those using any of
> these. The code for this functions is in the attached file. There is,
> however, one important difference to note. In the case the reference
> track efficiency could not be determined, the function _does_not_ return
> efficiency of 1.
> 
> The other differences in my function relates to particular treatment of
> the 12 degrees 1692 settings, i.e. both A and B polarity. The statistics
> is too low to require a count of 6 or more for each efficiency bin in T4
> and T5. Thus, so far, I require 5 counts or more. Also, what I've seen
> in the EfficiencyCalculator and BdstFsAna analogies is that they allow
> use of efficiency found in the underrun and overrun bins of the
> efficiency histograms. My function doesn't.
> 
> (Hope you follow so far. If not get some more coffee. ;) )
> 
> Next, I run over the available DSTs from 2001 AuAu. For each FS track I
> calculate the tracking detectors' efficiencies in terms of centrality
> and the tracks' x position and slope in each detector, again in close
> analogy with what's done in the before mentioned classes. I then fill
> histograms with the distribution of each tracking detector's efficiency.
> I've limited myself to the 10% most central events.
> 
> These distribution histograms are at the core of my puzzle. The reason
> is the great number of entries with efficiency of 1. I.e in the
> "eff_dist_<setting>" histograms in [1] they show up with large counts in
> the 100% bin on the far right. (Note the log scale.) Looking in
> particular at the settings with a lot of statistics, e.g. at 3 and 4
> degrees there seems to be nice (more or less) continuous distributions
> of tracking efficiency with peaks in the 80 or 90% range. In addition
> the bin for 100% is filled and clearly separated from the "continuum".
> 
> To me this indicates that tracking efficiency of 100% is _wrong_ and
> should _not_be_used_. IS THIS THE CASE?
> 
> (More coffee anyone?)
> 
> I then try to construct a combined tracking efficiency multiplying the
> individual efficiencies of T1, T2, T4 and T5 in full FS mode and only T1
> and T2 in FFS mode. I've done this in 2 ways. One is to make profile
> histograms for each detector and multiply these histograms together. I
> refer to this method as method 1. The "method1_<setting>" histograms in
> [1] are _not_ filled if the return value of GetDetectorEfficiency() is
> 1. (The FFS histograms for 20 and 30 degrees comes from parameterizing
> Truls' simulation results, and thus T1 and T2 histograms are empty.)
> 
> Also I try a second method, method 2, in which I calculate the
> efficiency while looping over the tracks in the DSTs. In the
> "method2_<setting>" figures in [1] the upper panel is a profile requiring
> (t1Eff != 1.) && (t2Eff != 1.) && (t4Eff != 1.) && (t5Eff != 1.)
> in FS mode and
> (t1Eff != 1.) && (t2Eff != 1.)
> in FFS mode for the histogram to be filled. The profile in the middle
> panel is filled _including_ individual efficiencies of 1. And the lower
> panel is simply the momentum distribution of the particles in the setting.
> 
> Method 2 causes lower statistics in the FS tracking efficiency profile,
> and therefore does not work well e.g. at the low statistics sample at 12
> degrees and magnet current of 1692. But where there is good statistics
> it is in fairly good agreement with method 1.
> 
> Note however that some FS settings show tracking efficiency of 30 and
> 40%, e.g. 8 degrees at A676 and 3 degrees at A1124. It is consistently
> smaller than what I've seen for tracking efficiency in other analyzes.
> Also the shape as function of momentum seems peculiar. Is there any
> reason for our efficiency to be decrease with increasing momentum which
> looks like being the case for many settings?
> 
> I therefore wonder if anyone has some feed back on this issue. In
> particular, is tracking detector efficiency of 1 a default and/or
> erroneous value which should be omitted when parameterizing the
> efficiency e.g. as function of momentum.
> 
> Best wishes from
> Jens Ivar
> 
> [1] http://www.ift.uib.no/~jens/trackingefficiency
> 
> 

-- 
:-) --------------------------- )-:
 Peter H L Christiansen
 pchristi@nbi.dk / (+41)764870425
:-D --------------------------- \-:



_______________________________________________
Brahms-dev-l mailing list
Brahms-dev-l@lists.bnl.gov
http://lists.bnl.gov/mailman/listinfo/brahms-dev-l