---------- Forwarded message ----------
Date: Wed, 28 Nov 2012 11:48:11 +0000
From: Marek Gazdzicki <Marek.Gazdzicki@cern.ch>
To: Nikolaos Davis <Nikolaos.Davis@cern.ch>,
     Herbert Stroebele <stroebel@ikf.uni-frankfurt.de>
Cc: Peter Seyboth <pxs@mppmu.mpg.de>,
     "fdiakono@phys.uoa.gr" <fdiakono@phys.uoa.gr>,
     Nikos Antoniou <nantonio@phys.uoa.gr>,
     Stroebele Herbert <stroebele@ikf.uni-frankfurt.de>
Subject: comments on proton intermittency analysis


   Dear All,

   I could not stay to the end of the last EVO meeting,
   sorry for this.
   In order to accelerate convergence to the final procedure/results
   I summarize below my suggestions.

  
--------------------------------------------------------------------------------
   A. First I would propose to obtain results using a "minimal"
procedure, namely:

   1. calculate  form the data for a single (any) position of bins:

                              {  DeltaF2(1), ... DeltaF2(M)  }

   (for simplicity in the brackets I put only index of a given binning
in pT space):

   2. obtain the covariance matrix of  {  DeltaF2(1), ... DeltaF2(M)  }
using the
       bootstrap method:

                         cov(  DeltaF2(1), ... DeltaF2(M) )

   3. the phi2 parameter is fitted to

                         {  DeltaF2(1), ... DeltaF2(M) }

       using the covariance matrix

                          cov[ DeltaF2(1), ... DeltaF2(M) ]

        and the chisq minimization method and the final result is obtained:

                         phi2 +/- sigma[phi2]


  
----------------------------------------------------------------------------------------------
    B. The "minimal" procedure is relatively simple, but still for many
a very complicated one,
        it is also only an approximate procedure. Therefore I propose to
test it using CMC, namely:

    1. set realistic input parameters to CMC, in particular phi2_MC
and   fraction of uncorrelated protons

    2. generate 10  MCdata samples

    3. analyze the MCdata samples using the "minimal" procedure and  get:

           { phi2_1 +/- sigma[phi2_1], phi2_2  +/- sigma[phi2_2]   ,
....   phi2_10  +/- sigma[phi2_10] }

    4. calculate <phi2>  and rms[phi2]  for the phi2 values and check
whether:

      4.1:    phi2_MC  \approx  <phi2>  ( within   rms[phi2]/sqrt(9) )

      4.2:  <sigma(phi2>  \approx  rms[phi2]


   C.  if needed
       (e.g. the "minimal" procedure gives:
        -  significantly biased results  (i.e. 4.1 and/or 4.2 are not
fulfilled),  and/or
        -  sigma(phi2) is too large )
        execute A. and B. for another procedure (.e.g.  with the
averaging over different bin positions)
        and check whether it yields results better than the "minimal"
procedure.

    Looking forward for your critical comments/discussion

    Best regards   Marek

   repeat  A. and B. for more