08-mar-2007 NvE AliSample extended with Median and Spread calculation for 1D histograms.

diff --git a/RALICE/AliSample.cxx b/RALICE/AliSample.cxx
index 174d011..4845dd0 100644 (file)
--- a/RALICE/AliSample.cxx
+++ b/RALICE/AliSample.cxx
@@ -626,7 +626,8 @@ void AliSample::SetStoreMode(Int_t mode)
 // storage of entered data is not needed. This is the default mode
 // of operation and is the most efficient w.r.t. cpu time and memory.
 // However, when calculation of a median, minimum or maximum is required,
-// then the data storage mode has be activated.
+// then the data storage mode has be activated, unless the statistics
+// are obtained from a specified input histogram.
 //
 // Note : Activation of storage mode can only be performed before the
 //        first data item is entered. 
@@ -651,6 +652,10 @@ Float_t AliSample::GetMedian(Int_t i)
 {
 // Provide the median of a certain variable.
 // For this functionality the storage mode has to be activated.
+//
+// Note : For large datasets it is more efficient to determine the median
+//        via the specification of a histogram. 
+//        See the other GetMedian memberfunction for details.
 
  if (fDim < i)
  {
@@ -741,6 +746,10 @@ Float_t AliSample::GetSpread(Int_t i)
 // Provide the spread w.r.t. the median of a certain variable.
 // The spread is defined as the average of |median-val(i)|.
 // For this functionality the storage mode has to be activated.
+//
+// Note : For large datasets it is more efficient to determine the spread
+//        via the specification of a histogram. 
+//        See the other GetSpread memberfunction for details.
 
  if (fDim < i)
  {
@@ -845,3 +854,79 @@ Float_t AliSample::GetMaximum(Int_t i) const
  return max;
 }
 ///////////////////////////////////////////////////////////////////////////
+Double_t AliSample::GetMedian(TH1* histo,Int_t mode) const
+{
+// Provide the median from the specified 1D histogram.
+// For this functionality it is not needed to activate the storage mode.
+//
+// This facility uses TH1::GetQuantiles to determine the median, which
+// provides a median value which in general is different from any of the
+// central bin X values. The user may force the returned median to be
+// the corresponding central bin X value via the "mode" input argument.
+//
+// mode = 0 ==> The pure TH1::GetQuantiles median value is returned
+//        1 ==> The corresponding central bin X value is returned as median
+//
+// By default mode=1 will be used, to agree with the AliSample philosophy.
+
+ if (!histo) return 0;
+
+ Double_t median=0;
+
+ Double_t q[1];
+ Double_t p[1]={0.5};
+ Double_t integral=histo->ComputeIntegral();
+ Int_t nq=histo->GetQuantiles(1,q,p);
+
+ if (!nq) return 0;
+
+ median=q[0];
+ if (mode)
+ {
+  Int_t mbin=histo->FindBin(q[0]);
+  median=histo->GetBinCenter(mbin);
+ }
+
+ return median;
+}
+///////////////////////////////////////////////////////////////////////////
+Double_t AliSample::GetSpread(TH1* histo,Int_t mode) const
+{
+// Provide the spread w.r.t. the median for the specified 1D histogram.
+// The spread is defined as the average of |median-xval|.
+// For this functionality it is not needed to activate the storage mode.
+//
+// This facility uses TH1::GetQuantiles to determine the median, which
+// provides a median value which in general is different from any of the
+// central bin X values. The user may force the used median to be
+// the corresponding central bin X value via the "mode" input argument.
+//
+// mode = 0 ==> The pure TH1::GetQuantiles median value is used
+//        1 ==> The corresponding central bin X value is used as median
+//
+// By default mode=1 will be used, to agree with the AliSample philosophy.
+
+ if (!histo) return 0;
+
+ Double_t spread=0;
+
+ Double_t median=GetMedian(histo,mode);
+
+ Int_t nbins=histo->GetNbinsX();
+ Double_t x=0,y=0,ysum=0;
+ for (Int_t jbin=1; jbin<=nbins; jbin++)
+ {
+  x=histo->GetBinCenter(jbin);
+  y=histo->GetBinContent(jbin);
+  if (y>0)
+  {
+   spread+=fabs(x-median)*y;
+   ysum+=y;
+  }
+ }
+
+ if (ysum>0) spread=spread/ysum;
+
+ return spread;
+}
+///////////////////////////////////////////////////////////////////////////

diff --git a/RALICE/AliSample.h b/RALICE/AliSample.h
index d6924b1..f157427 100644 (file)
--- a/RALICE/AliSample.h
+++ b/RALICE/AliSample.h
@@ -9,6 +9,7 @@
 
 #include "Rtypes.h"
 #include "TArrayF.h"
+#include "TH1.h"
  
 class AliSample
 {
@@ -39,6 +40,8 @@ class AliSample
   void Data(Int_t i, Int_t j) const;            // Stat. info for i-th and j-th variable
   void SetStoreMode(Int_t mode=1);              // Set mode for storage of all entered data
   Int_t GetStoreMode() const;                   // Provide storage mode of all entered data
+  Double_t GetMedian(TH1* histo,Int_t mode=1) const; // Provide median from the specified 1D histogram
+  Double_t GetSpread(TH1* histo,Int_t mode=1) const; // Provide spread w.r.t. the median from the 1D histo
  
  private:
   Int_t fDim;                      // Dimension of the sample

diff --git a/RALICE/history.txt b/RALICE/history.txt
index 15dc1b5..c16f173 100644 (file)
--- a/RALICE/history.txt
+++ b/RALICE/history.txt
@@ -749,4 +749,5 @@
 08-dec-2006 NvE Memberfunctions SetUT introduced in AliTimestamp to provide convenient
                 ways to set a specific UT date/time.
 20-jan-2007 NvE AliMath extended with Nfac(), LnNfac() and LogNfac() facilities.
+08-mar-2007 NvE AliSample extended with Median and Spread calculation for 1D histograms.