benhoob/HWW/correlationscatters.C

#include "tmvaglob.C"

// this macro plots the correlations (as scatter plots) of
// the various input variable combinations used in TMVA (e.g. running
// TMVAnalysis.C).  Signal and Background are plotted separately

// input: - Input file (result from TMVA),
//        - normal/decorrelated/PCA
//        - use of TMVA plotting TStyle
void correlationscatters( TString fin = "TMVA.root", TString var= "var3", 
                          TString dirName_ = "InputVariables_Id", TString title = "TMVA Input Variable",
                          Bool_t isRegression = kFALSE,
                          Bool_t useTMVAStyle = kTRUE )
{
   // set style and remove existing canvas'
   TMVAGlob::Initialize( useTMVAStyle );

   TString extension = dirName_;
   extension.ReplaceAll( "InputVariables", "" );
   extension.ReplaceAll( " ", "" );
   if (extension == "") extension = "_Id"; // use 'Id' for 'idendtity transform'

   var.ReplaceAll( extension, "" );
   cout << "Called macro \"correlationscatters\" for variable: \"" << var 
        << "\", transformation type \"" << dirName_ 
        << "\" (extension: \"" << extension << "\")" << endl;

   // checks if file with name "fin" is already open, and if not opens one
   TFile* file = TMVAGlob::OpenFile( fin );  

   TString dirName = dirName_ + "/CorrelationPlots";
  
   // find out number of input variables   
   TDirectory* vardir = (TDirectory*)gDirectory->Get( "InputVariables_Id" );
   if (!vardir) {
      cout << "ERROR: no such directory: \"InputVariables\"" << endl;
      return;
   }
   Int_t noVars = TMVAGlob::GetNumberOfInputVariables( vardir ); // subtraction of target(s) no longer necessary

   TDirectory* dir = (TDirectory*)gDirectory->Get( dirName );
   if (dir==0) {
      cout << "No information about " << extension << " available in " << fin << endl;
      return;
   }
   dir->cd();

   TListIter keyIt(dir->GetListOfKeys());
   TKey* key = 0;
   Int_t noPlots = noVars - 1;
   
   cout << "noPlots: " << noPlots << " --> noVars: " << noVars << endl;
   if (noVars != Int_t(noVars)) {
      cout << "*** Warning: problem in inferred number of variables ... not an integer *** " << endl;
   }

   // define Canvas layout here!
   // default setting
   Int_t xPad;  // no of plots in x
   Int_t yPad;  // no of plots in y
   Int_t width; // size of canvas
   Int_t height;
   switch (noPlots) {
   case 1:
      xPad = 1; yPad = 1; width = 400; height = width; break;
   case 2:
      xPad = 2; yPad = 1; width = 700; height = 0.55*width; break;
   case 3:
      xPad = 3; yPad = 1; width = 800; height = 0.4*width; break;
   case 4:
      xPad = 2; yPad = 2; width = 600; height = width; break;
   default:
      xPad = 3; yPad = 2; width = 800; height = 0.55*width; break;
   }
   Int_t noPadPerCanv = xPad * yPad ;   

   // counter variables
   Int_t countCanvas = 0;

   // loop over all objects in "input_variables" directory
   Int_t ncls = (isRegression ? 1 : 2);
   TString thename[2] = { "_Signal", "_Background" };
   if (isRegression) thename[0] = "_Regression";
   for (UInt_t itype = 0; itype < 2; itype++) {

      TIter next(gDirectory->GetListOfKeys());
      TKey   * key  = 0;
      TCanvas* canv = 0;

      Int_t countPad    = 0;
   
      while ((key = (TKey*)next())) {

         if (key->GetCycle() != 1) continue;

         // make sure, that we only look at histograms
         TClass *cl = gROOT->GetClass(key->GetClassName());
         if (!cl->InheritsFrom("TH1")) continue;
         TH1 *scat = (TH1*)key->ReadObj();
         TString hname = scat->GetName();

         // check for all signal histograms
         if (! (hname.EndsWith( thename[itype] + extension ) && 
                hname.Contains( TString("_") + var + "_" ) && hname.BeginsWith("scat_")) ) {
            scat->Delete();
            continue; 
         }

         // found a new signal plot
            
         // create new canvas
         if (countPad%noPadPerCanv==0) {
            ++countCanvas;
            TString ext = extension; ext.Remove( 0, 1 );
            canv = new TCanvas( Form("canvas%d", countCanvas), 
                                Form("Correlation profiles for '%s'-transformed %s variables", 
                                     ext.Data(), (isRegression ? "" : (itype==0) ? "signal" : "background")),
                                countCanvas*50+200, countCanvas*20, width, height ); 
            canv->Divide(xPad,yPad);
         }

         if (!canv) continue;

         canv->cd(countPad++%noPadPerCanv+1);

         // find the corredponding backgrouns histo
         TString bgname = hname;
         bgname.ReplaceAll("scat_","prof_");
         TH1 *prof = (TH1*)gDirectory->Get(bgname);
         if (prof == NULL) {
            cout << "ERROR!!! couldn't find background histo for" << hname << endl;
            exit;
         }
         // this is set but not stored during plot creation in MVA_Factory
         TMVAGlob::SetSignalAndBackgroundStyle( scat, prof );

         // chop off "signal" 
         TMVAGlob::SetFrameStyle( scat, 1.2 );

         // normalise both signal and background
         scat->Scale( 1.0/scat->GetSumOfWeights() );

         // finally plot and overlay       
         Float_t sc = 1.1;
         if (countPad==2) sc = 1.3;
         scat->SetMarkerColor(  4);
         scat->Draw("col");      
         prof->SetMarkerColor( TMVAGlob::UsePaperStyle ? 1 : 2  );
         prof->SetMarkerSize( 0.2 );
         prof->SetLineColor( TMVAGlob::UsePaperStyle ? 1 : 2 );
         prof->SetLineWidth( TMVAGlob::UsePaperStyle ? 2 : 1 );
         prof->SetFillStyle( 3002 );
         prof->SetFillColor( 46 );
         prof->Draw("samee1");
         // redraw axes
         scat->Draw("sameaxis");

         // save canvas to file
         if (countPad%noPadPerCanv==0) {
            canv->Update();

            TString fname = Form( "plots/correlationscatter_%s_%s_c%i",var.Data(), extension.Data(), countCanvas );
            TMVAGlob::plot_logo();
            TMVAGlob::imgconv( canv, fname );
         }
      }
      if (countPad%noPadPerCanv!=0) {
         canv->Update();

         TString fname = Form( "plots/correlationscatter_%s_%s_c%i",var.Data(), extension.Data(), countCanvas );
         TMVAGlob::plot_logo();
         TMVAGlob::imgconv( canv, fname );
      }
   }
}
Revision:	1.1
Committed:	Mon Feb 14 12:39:14 2011 UTC (14 years, 3 months ago) by benhoob
Content type:	text/plain
Branch:	MAIN
CVS Tags:	HEAD
Log Message:	Initial commit
#	User	Rev	Content
1	benhoob	1.1	#include "tmvaglob.C"
2
3			// this macro plots the correlations (as scatter plots) of
4			// the various input variable combinations used in TMVA (e.g. running
5			// TMVAnalysis.C). Signal and Background are plotted separately
6
7			// input: - Input file (result from TMVA),
8			// - normal/decorrelated/PCA
9			// - use of TMVA plotting TStyle
10			void correlationscatters( TString fin = "TMVA.root", TString var= "var3",
11			TString dirName_ = "InputVariables_Id", TString title = "TMVA Input Variable",
12			Bool_t isRegression = kFALSE,
13			Bool_t useTMVAStyle = kTRUE )
14			{
15			// set style and remove existing canvas'
16			TMVAGlob::Initialize( useTMVAStyle );
17
18			TString extension = dirName_;
19			extension.ReplaceAll( "InputVariables", "" );
20			extension.ReplaceAll( " ", "" );
21			if (extension == "") extension = "_Id"; // use 'Id' for 'idendtity transform'
22
23			var.ReplaceAll( extension, "" );
24			cout << "Called macro \"correlationscatters\" for variable: \"" << var
25			<< "\", transformation type \"" << dirName_
26			<< "\" (extension: \"" << extension << "\")" << endl;
27
28			// checks if file with name "fin" is already open, and if not opens one
29			TFile* file = TMVAGlob::OpenFile( fin );
30
31			TString dirName = dirName_ + "/CorrelationPlots";
32
33			// find out number of input variables
34			TDirectory* vardir = (TDirectory*)gDirectory->Get( "InputVariables_Id" );
35			if (!vardir) {
36			cout << "ERROR: no such directory: \"InputVariables\"" << endl;
37			return;
38			}
39			Int_t noVars = TMVAGlob::GetNumberOfInputVariables( vardir ); // subtraction of target(s) no longer necessary
40
41			TDirectory* dir = (TDirectory*)gDirectory->Get( dirName );
42			if (dir==0) {
43			cout << "No information about " << extension << " available in " << fin << endl;
44			return;
45			}
46			dir->cd();
47
48			TListIter keyIt(dir->GetListOfKeys());
49			TKey* key = 0;
50			Int_t noPlots = noVars - 1;
51
52			cout << "noPlots: " << noPlots << " --> noVars: " << noVars << endl;
53			if (noVars != Int_t(noVars)) {
54			cout << "* Warning: problem in inferred number of variables ... not an integer * " << endl;
55			}
56
57			// define Canvas layout here!
58			// default setting
59			Int_t xPad; // no of plots in x
60			Int_t yPad; // no of plots in y
61			Int_t width; // size of canvas
62			Int_t height;
63			switch (noPlots) {
64			case 1:
65			xPad = 1; yPad = 1; width = 400; height = width; break;
66			case 2:
67			xPad = 2; yPad = 1; width = 700; height = 0.55*width; break;
68			case 3:
69			xPad = 3; yPad = 1; width = 800; height = 0.4*width; break;
70			case 4:
71			xPad = 2; yPad = 2; width = 600; height = width; break;
72			default:
73			xPad = 3; yPad = 2; width = 800; height = 0.55*width; break;
74			}
75			Int_t noPadPerCanv = xPad * yPad ;
76
77			// counter variables
78			Int_t countCanvas = 0;
79
80			// loop over all objects in "input_variables" directory
81			Int_t ncls = (isRegression ? 1 : 2);
82			TString thename[2] = { "_Signal", "_Background" };
83			if (isRegression) thename[0] = "_Regression";
84			for (UInt_t itype = 0; itype < 2; itype++) {
85
86			TIter next(gDirectory->GetListOfKeys());
87			TKey * key = 0;
88			TCanvas* canv = 0;
89
90			Int_t countPad = 0;
91
92			while ((key = (TKey*)next())) {
93
94			if (key->GetCycle() != 1) continue;
95
96			// make sure, that we only look at histograms
97			TClass *cl = gROOT->GetClass(key->GetClassName());
98			if (!cl->InheritsFrom("TH1")) continue;
99			TH1 scat = (TH1)key->ReadObj();
100			TString hname = scat->GetName();
101
102			// check for all signal histograms
103			if (! (hname.EndsWith( thename[itype] + extension ) &&
104			hname.Contains( TString("_") + var + "_" ) && hname.BeginsWith("scat_")) ) {
105			scat->Delete();
106			continue;
107			}
108
109			// found a new signal plot
110
111			// create new canvas
112			if (countPad%noPadPerCanv==0) {
113			++countCanvas;
114			TString ext = extension; ext.Remove( 0, 1 );
115			canv = new TCanvas( Form("canvas%d", countCanvas),
116			Form("Correlation profiles for '%s'-transformed %s variables",
117			ext.Data(), (isRegression ? "" : (itype==0) ? "signal" : "background")),
118			countCanvas50+200, countCanvas20, width, height );
119			canv->Divide(xPad,yPad);
120			}
121
122			if (!canv) continue;
123
124			canv->cd(countPad++%noPadPerCanv+1);
125
126			// find the corredponding backgrouns histo
127			TString bgname = hname;
128			bgname.ReplaceAll("scat_","prof_");
129			TH1 prof = (TH1)gDirectory->Get(bgname);
130			if (prof == NULL) {
131			cout << "ERROR!!! couldn't find background histo for" << hname << endl;
132			exit;
133			}
134			// this is set but not stored during plot creation in MVA_Factory
135			TMVAGlob::SetSignalAndBackgroundStyle( scat, prof );
136
137			// chop off "signal"
138			TMVAGlob::SetFrameStyle( scat, 1.2 );
139
140			// normalise both signal and background
141			scat->Scale( 1.0/scat->GetSumOfWeights() );
142
143			// finally plot and overlay
144			Float_t sc = 1.1;
145			if (countPad==2) sc = 1.3;
146			scat->SetMarkerColor( 4);
147			scat->Draw("col");
148			prof->SetMarkerColor( TMVAGlob::UsePaperStyle ? 1 : 2 );
149			prof->SetMarkerSize( 0.2 );
150			prof->SetLineColor( TMVAGlob::UsePaperStyle ? 1 : 2 );
151			prof->SetLineWidth( TMVAGlob::UsePaperStyle ? 2 : 1 );
152			prof->SetFillStyle( 3002 );
153			prof->SetFillColor( 46 );
154			prof->Draw("samee1");
155			// redraw axes
156			scat->Draw("sameaxis");
157
158			// save canvas to file
159			if (countPad%noPadPerCanv==0) {
160			canv->Update();
161
162			TString fname = Form( "plots/correlationscatter_%s_%s_c%i",var.Data(), extension.Data(), countCanvas );
163			TMVAGlob::plot_logo();
164			TMVAGlob::imgconv( canv, fname );
165			}
166			}
167			if (countPad%noPadPerCanv!=0) {
168			canv->Update();
169
170			TString fname = Form( "plots/correlationscatter_%s_%s_c%i",var.Data(), extension.Data(), countCanvas );
171			TMVAGlob::plot_logo();
172			TMVAGlob::imgconv( canv, fname );
173			}
174			}
175			}