Questions to ruben.shahoyan@cern.ch

------------------------------------------------------------
PbPb dN/dEta analysis can be done in 2 ways:

1) For single centrality bin:
Using AliTrackletTaskUni.{h,cxx} steered by the
runAAF.C and MyAnalysisMacro.C

It will produce a file with 3D (sparse) histos "delta" vs Zv vs eta
where delta is either "weigted distance" and/or "dphi - bend", which is used
both the match the tails and to define the signal cut.
The exact value of cut can be decided at the stage of processing the histos,
so different signal and tail matching thresholds can be tested 

Note that wide Zv eta range can be processed and then restricted at 
the correction stage.

One has to run runAAF.C for over the data and MC datasets and then analyse them
using the CorrectSpectra.C macro (the cuts/variables use... must be set 
beforehand)


2) For multiple centrality bins at once:
AliTrackletTaskMulti.{h,cxx} steered by the 
runAAFMulti.C and MyAnalysisMacroTrackletMulti.C

One should set the in the AliTrackletTaskMulti the
const Float_t  AliTrackletTaskMulti::fgkCentPerc[] ... aray with definition of selected 
centrality bins, i.e. {0.,5.,10.,100} will creat bin0 for 0-5% centrality, bin1 for 5-10% etc.

The variable on which centrality is defined is selected via runAAFMulti useCentVar parameter,
should correspond to one of AliTrackletMultTask 
enum {kCentV0M,kCentFMD,kCentTRK,kCentTKL,kCentCL0,kCentCL1,kCentV0MvsFMD,kCentTKLvsV0,kCentZEMvsZDC,kNCentTypes}; 

The produces for each centrality bin a set of histos, particularly 2D histos for
Zv vs eta with the signal cut on "distance" already applied (by defaults one for 
"w.dist" another for "dphi-bend" + 1D histo of "distance" for selected Zv,eta
range (to be used for the bg matching the tails of data"

Note: Zv, eta ranges must be defined at data processing stage, as well as the 
signal cut (cutSigNStd in the runAAFMulti) and number of st.dev to keep (nStdDev)

One has to runAAFMulti.C for over the data and MC datasets and then analyse them
using the CorrectSpectraMulti.C macro 

The typical use of runAAFMulti is (used for PbPb analysis in 2010)
root -q 'runAAFMulti.C(
"/alice/data/LHC10h_000137366_p2",                                    // input 
"resMultiWide_LHC10h_000137366_p2_eta_m08_p08_zv_m7_p7_zdczem.root",  // output
-1,        // N events to process (-1 : all events)
-0.5,0.5,  // eta selection
-7,7,      // Zv selection
8,         // centrality variable
0.7520,    // rescale MC V0 to match data
 1.5,      // cut on weighed distance used to extract signal
 -1,       // cut on dPhi-phiBent used to extract signal (if negative -> dphi*sqrt(cutSigNStd), recommended!)
kTRUE,     // fill MC info (macro detects automatically that MC is analysed from ../sim/.. in the input dataset name
kTRUE,     // redo tracklets reconstruction and use new AliMuliplicity to fill histos
kTRUE      // generate injected bg
... the rest is better to not touch..
)'

--------------------------------
Both methods can use 3 types of generated bg: injection, rotation and mixing.
Simultaneous eployment of all these methods is also possible, but may create 
a memory problem.

The corresponding CorrectSpectra... macros must be tuned for the bg.type used.

-----------------------------------------------------
Update: Wed Apr 20 16:23:55 CEST 2011
Addapted AliTrackletTaskMulti for pp data analysis w/o bg generation. In this case 
the recpoints and the connection to OCDB (alien libs) are not needed.
One should set the 
const Float_t  AliTrackletTaskMulti::fgkCentPerc[] = {0,100};
and chose a centrality variable availabe in pp, like V0.
Note that in this mode the AliMultiplicity object from the input dataset is used.

The typical call of runAAFMulti is:

root -q 'runAAFMulti.C("/alice/data/LHC10e_000130844_p2","resppWide_LHC10e_130844_p2_eta_m26_p26_zv_m20_p20a.root",
-1,
-2.6,2.6,
-20,20,0, 
0.7520,
2.,        // put it to large value to not affect the tracklet selection from existing AliMultiplicity object
-1,
kTRUE,
kFALSE,   // DO NOT do new reco of tracklets
kFALSE,   // DO NOT do bg. generation by injection
kFALSE,   // DO NOT do bg.generation by rotation
kFALSE,   // DO NOT do bg.generation by mixing
3.14159e+00, // irrelevant
1.,          // irrelevant 
kFALSE,    // NO scaling of dtheta by sin^2(theta) (that's how pp data was reconstructed so far...)
2,         // irrelevant
0.08,      // dphi tolerance
0.025      // dtheta tolerance
)'

The sample macro ppcor.C shows how to extract dNdEta from the outputs of data and MC.

----------
Thu Apr 21 23:27:19 CEST 2011
Put the version for pp in the AliTrackletTaskMultipp.h/cxx (rename it to AliTrackletTaskMulti.h/cxx 
before running runAAFMulti as mentioned above.
Added additional method 
FillClusterAutoCorrelationFromMult(const AliMultiplicity* mlt, double zVertex)
to fill the inter-cluster distances (all to all and used to unused) for clusters
vs Z and Phi
------------------------------------------------------------