Re: BB vertex offset

From: Christian Holm Christensen (cholm@hehi03.nbi.dk)
Date: Mon Apr 16 2001 - 08:00:40 EDT

  • Next message: Stephen J. Sanders: "Re: BB vertex offset"

    Hi Steve et al, 
    
    On Thu, 12 Apr 2001 22:15:57 -0500
    "Stephen J. Sanders" <ssanders@ukans.edu> wrote
    concerning ": BB vertex offset":
    > Hi,
    > one significant difference in the two calculations to the approx. 2 cm
    > difference found between the BB and TPM1 vertices.  
    
    Bjorn, Yury, do you know why there's this difference? 
    
    Steve, can we be absolutely sure that the BB vertex is "off" by 2 cm.?
    Or the ZDC vertex for that matter? Couldn't it be that the TPM1 vertex
    is off by 2 cm. (or 4 or 6 or ...)? In short, what is it that makes
    everyone so sure that the TPM1 vertex is the "best" and should always
    be right?  I've previously suggested trying to extract a primary
    vertex position from all avaliable primary vertex meassuearments in a
    given event. How say you (all) on that? 
    
    Bjorn, I've asked you this before but I can't remember getting a
    reply: What's the data members BrVertex::fVertex_sigma_<x|y> and
    BrVertex::fVariance<X|Y>? Sigma (or standard diviation or square root
    of the variance) is an intrinsic property of some probablity
    distribution, while "sample variance" (s^2) and "sample standard
    diviation" (s) is meassured quantaties. If, what you mean by these two
    members is that fVertex_sigma_<x|y> = sqrt(fVariance<X|Y>), then one
    of the sets is redundant (fVertex_sigma_<x|y>) and should not be
    present. A method (GetDiviationX() const { return
    TMath::Sqrt(fVarianceX); }) is what you need. 
    
    > In the "old" analysis, I had explicitly worked this into the vertex
    > I  (illegally) passed to the BrRdoModuleMult class.   However,
    > putting such an artificial offset into the BrSiRdoMult and
    > BrTileRdoTile classes (both of which need the vertex) is likely to
    > cause consternation in the future. 
    
    Oh yes. Another reason to use all avaliable vertex meassuarements to
    try to find some "common" vertex, stored in the event node.  
    
    > Is there a reason why this offset is not incorporated in the vertex
    > returned by the BrRdoBB module?
    
    To me, it seems that such a "correction" should not be imposed on the
    meassured data, at least not at that stage. 
    
    > For the moment, I have updated the mult modules to incorporate the
    > vertex offset.  
    
    It should have a temporary character, maybe behind preprocessor
    guards. 
    
    Uh. Steve (and others) please don't use fabs, sin, cos, tan, and other
    functions from cmath. Use TMath/BrMath instead. TMath is there to
    insure portabliblity (various compilers have different ideas of
    ANSI/ISO standard C - yes I know, stupid - ROOT doesn't). 
    
    On Sun, 15 Apr 2001 21:43:11 -0500
    "Stephen J. Sanders" <ssanders@ukans.edu> wrote
    concerning ": mult changes":
    > Hi,  The BrSiRdoModule and BrTileRdoModule classes now report the same
    > (to better than .05%) multiplicities as
    > BrRdoModuleMult  ---IF--- fOutlierMethod is not set, or is set to
    > kNoCorrection.  
    
    Uh. That's a bit odd. To me it sounds like your stuff in
    BrRdoModuleMult to correct for exesive energy events did diddly
    squad - that can't be right.
     
    > The DEFAULT in BrTileRdoModule is now to set fOutlierMethod to
    > kNBICorrection.  
    
    Uh. I really don't like that name. The correction that is conditional
    on fOutlierCorrection == kNBICorrection, is a correction to the number
    of possible tile hit to the actual number of tiles hit so that
    multiplicities across different events are comparable. Hence, the
    name should rather be something like kHitCorrection. Descriptive
    names, please.
    
    > This returns the previous mult for this class so as to maintain
    > consistentency with the current NBI centrality calculation.
    
    Yes, setting the outlier correction method to something different
    could screw up the centrality determination, as done in
    BrTileCentModule. Again, this is a very good reason to choose some
    other meassure for the "centrality" of an event, rather then the, much
    corrected, multiplicity.  What do you think? 
    
    Yours, 
    
    Christian  -----------------------------------------------------------
    Holm Christensen                             Phone:  (+45) 35 35 96 91 
      Sankt Hansgade 23, 1. th.                  Office: (+45) 353  25 305 
      DK-2200 Copenhagen N                       Web:    www.nbi.dk/~cholm    
      Denmark                                    Email:       cholm@nbi.dk
    



    This archive was generated by hypermail 2b29 : Mon Apr 16 2001 - 08:02:35 EDT