| 1 |
//root
|
| 2 |
#include "TCanvas.h"
|
| 3 |
#include "TPostScript.h"
|
| 4 |
#include "TH1D.h"
|
| 5 |
#include "TF1.h"
|
| 6 |
#include "TLegend.h"
|
| 7 |
#include "TStyle.h"
|
| 8 |
#include "TRandom3.h"
|
| 9 |
#include "TGraph.h"
|
| 10 |
#include "TLatex.h"
|
| 11 |
#include "TLine.h"
|
| 12 |
#include "TArrow.h"
|
| 13 |
#include "TLimit.h"
|
| 14 |
#include "TLimitDataSource.h"
|
| 15 |
#include "TConfidenceLevel.h"
|
| 16 |
#include "TVectorT.h"
|
| 17 |
|
| 18 |
//User
|
| 19 |
#include "cls.h"
|
| 20 |
#include "solve.h"
|
| 21 |
#include "table.h"
|
| 22 |
#include "significance.h"
|
| 23 |
#include "ConfigFile.h"
|
| 24 |
//system
|
| 25 |
#include <iostream>
|
| 26 |
#include <iomanip>
|
| 27 |
#include <cmath>
|
| 28 |
#include <assert.h>
|
| 29 |
#include <fstream>
|
| 30 |
#include <time.h>
|
| 31 |
|
| 32 |
cls::cls():
|
| 33 |
plotindex_(0),fNMC_(5000),outputfilename_("cls.ps"),signal_(0),backgd_(0),data_(0)
|
| 34 |
{
|
| 35 |
stat_=false;syst_=false;
|
| 36 |
do1DScan_ = false;
|
| 37 |
if (!data_ && signal_ && backgd_) {
|
| 38 |
data_ = (TH1D*)MakePseudoData( backgd_ );//, signal_ );
|
| 39 |
isPseudoData = true;
|
| 40 |
} else
|
| 41 |
isPseudoData = false;
|
| 42 |
}
|
| 43 |
|
| 44 |
cls::cls(std::string n,TH1*s,TH1*b,TH1*d):
|
| 45 |
plotindex_(0),fNMC_(50000),outputfilename_(n),signal_(s),backgd_(b),data_(d)
|
| 46 |
{
|
| 47 |
stat_=false;syst_=false;
|
| 48 |
do1DScan_ = false;
|
| 49 |
if (!data_ && signal_ && backgd_) {
|
| 50 |
data_ = (TH1D*)MakePseudoData( backgd_ );//, signal_ );
|
| 51 |
isPseudoData = true;
|
| 52 |
} else
|
| 53 |
isPseudoData = false;
|
| 54 |
}
|
| 55 |
|
| 56 |
cls::cls(std::string n,std::string ScanParName,std::vector<double>ScanPar,std::vector<TH1*>s,TH1*b,TH1*d):
|
| 57 |
plotindex_(0),fNMC_(100000),outputfilename_(n),ScanParName_(ScanParName),ScanPar_(ScanPar),backgd_(b),data_(d)
|
| 58 |
{
|
| 59 |
stat_=true;syst_=false;
|
| 60 |
signals_ = s;
|
| 61 |
signal_ = (signals_.size()>0 ? signals_[0] : 0);
|
| 62 |
do1DScan_ = (signals_.size()>1 ? true : false);
|
| 63 |
|
| 64 |
if (!data_ && signal_ && backgd_) {
|
| 65 |
data_ = (TH1D*)MakePseudoData( backgd_ );//, signal_ );
|
| 66 |
isPseudoData = true;
|
| 67 |
} else
|
| 68 |
isPseudoData = false;
|
| 69 |
}
|
| 70 |
|
| 71 |
|
| 72 |
void cls::GetXandYbyInterpolation( const double x, TH1D *& sig)
|
| 73 |
{
|
| 74 |
///naive linear interpolation for each bin contents and bin error:
|
| 75 |
sig=0;
|
| 76 |
double xmin=-99999, xmax=999999;
|
| 77 |
int imin, imax;
|
| 78 |
for (std::vector<double>::const_iterator it=ScanPar_.begin();it!=ScanPar_.end();++it){
|
| 79 |
if (*it<=x && xmin<*it) { xmin= *it; imin=it-ScanPar_.begin(); continue;}
|
| 80 |
if (*it>=x && xmax>*it) { xmax= *it; imax=it-ScanPar_.begin(); }
|
| 81 |
}
|
| 82 |
if (xmin<0) return;
|
| 83 |
sig = (TH1D*)signals_[imin]->Clone();
|
| 84 |
if (xmax>99999||xmin==xmax) return;
|
| 85 |
TH1D * SIG = (TH1D*)signals_[imax]->Clone();
|
| 86 |
|
| 87 |
for (int i=1; i<=sig->GetXaxis()->GetNbins(); ++i){//avoid under- and overflow bins
|
| 88 |
//std::cout <<"sig_left: x="<<xmin<<", y="<<sig->GetBinContent(i);
|
| 89 |
double k = (SIG->GetBinContent(i)-sig->GetBinContent(i))/(xmax-xmin);
|
| 90 |
double kerr = (SIG->GetBinError(i)-sig->GetBinError(i))/(xmax-xmin);
|
| 91 |
sig->SetBinContent(i,sig->GetBinContent(i)+k*(x-xmin));
|
| 92 |
sig->SetBinError(i,sig->GetBinError(i)+kerr*(x-xmin));
|
| 93 |
//std::cout <<"; sig_new: x="<<x<<", y="<<sig->GetBinContent(i)
|
| 94 |
// <<"; sig_right: x="<<xmax<<", y="<<SIG->GetBinContent(i)
|
| 95 |
// <<std::endl;
|
| 96 |
}
|
| 97 |
delete SIG;
|
| 98 |
}
|
| 99 |
|
| 100 |
|
| 101 |
|
| 102 |
double cls::operator()(const double x, const double * par)
|
| 103 |
{
|
| 104 |
double cl;
|
| 105 |
TH1D *sig=0;
|
| 106 |
if (do1DScan_) {
|
| 107 |
GetXandYbyInterpolation(x, sig);
|
| 108 |
if (!sig) return 99999;
|
| 109 |
}
|
| 110 |
else {
|
| 111 |
sig = (TH1D*)signal_->Clone();
|
| 112 |
sig->Scale( x );
|
| 113 |
}
|
| 114 |
TLimitDataSource * source = new TLimitDataSource(sig,(TH1D*)backgd_,(TH1D*)data_);
|
| 115 |
TConfidenceLevel * confl = TLimit::ComputeLimit(source,fNMC_,stat_);
|
| 116 |
if (par[1]) cl = confl->CLs();
|
| 117 |
else cl = confl->GetExpectedCLs_b();
|
| 118 |
delete confl;
|
| 119 |
delete sig;
|
| 120 |
delete source;
|
| 121 |
return cl - par[0]; //par[0] is the requested Confidence Level (e.g. 0.05)
|
| 122 |
}
|
| 123 |
|
| 124 |
double cls::GetObservedXsecLimit(double cl, double min, double max)
|
| 125 |
{
|
| 126 |
double par[] = {cl,true};
|
| 127 |
return TSolve<cls>( this, &cls::operator(), par, min, max)();
|
| 128 |
}
|
| 129 |
double cls::GetExpectedXsecLimit(double cl, double min, double max)
|
| 130 |
{
|
| 131 |
double par[] = {cl,false};
|
| 132 |
return TSolve<cls>( this, &cls::operator(), par, min, max)();
|
| 133 |
}
|
| 134 |
|
| 135 |
TH1 * cls::MakePseudoData(TH1 const *b, TH1 const *s)
|
| 136 |
{
|
| 137 |
char * name = new char[100];
|
| 138 |
sprintf(name,"data%d",plotindex_++);
|
| 139 |
char * title = new char[128];
|
| 140 |
sprintf(title,";%s;events",b->GetXaxis()->GetTitle());
|
| 141 |
TH1 * result = (TH1D*)b->Clone( title );
|
| 142 |
TRandom3 * rand = new TRandom3(0);
|
| 143 |
for (int i=0; i<=b->GetNbinsX(); ++i){
|
| 144 |
double entry;
|
| 145 |
double uncert;
|
| 146 |
if (s==0) {
|
| 147 |
entry = rand->Poisson( b->GetBinContent(i) );
|
| 148 |
uncert = rand->Gaus(0, b->GetBinError(i));
|
| 149 |
}
|
| 150 |
else {
|
| 151 |
entry = rand->Poisson( b->GetBinContent(i)+s->GetBinContent(i) );
|
| 152 |
uncert = rand->Gaus(0, b->GetBinError(i)+s->GetBinError(i));
|
| 153 |
}
|
| 154 |
result->SetBinContent(i,entry+uncert);
|
| 155 |
result->SetBinError(i,sqrt(entry+uncert));
|
| 156 |
}
|
| 157 |
delete name;
|
| 158 |
delete title;
|
| 159 |
delete rand;
|
| 160 |
return result;
|
| 161 |
}
|
| 162 |
|
| 163 |
void cls::Draw(bool doeps)
|
| 164 |
{
|
| 165 |
if (do1DScan_) DrawVsSignalParam(doeps);
|
| 166 |
else DrawVsXsec(doeps);
|
| 167 |
}
|
| 168 |
|
| 169 |
void cls::Sort(double &v1, double &v2)
|
| 170 |
{
|
| 171 |
if (v1>v2){
|
| 172 |
double temp(v1);
|
| 173 |
v1=v2;
|
| 174 |
v2=temp;
|
| 175 |
}
|
| 176 |
}
|
| 177 |
|
| 178 |
void cls::DrawVsSignalParam(bool doeps)
|
| 179 |
{
|
| 180 |
gStyle->SetHistFillColor(0);
|
| 181 |
gStyle->SetPalette(1);
|
| 182 |
gStyle->SetCanvasColor(0);
|
| 183 |
gStyle->SetCanvasBorderMode(0);
|
| 184 |
gStyle->SetPadColor(0);
|
| 185 |
gStyle->SetPadBorderMode(0);
|
| 186 |
gStyle->SetFrameBorderMode(0);
|
| 187 |
|
| 188 |
gStyle->SetTitleFillColor(0);
|
| 189 |
gStyle->SetTitleBorderSize(0);
|
| 190 |
gStyle->SetTitleX(0.10);
|
| 191 |
gStyle->SetTitleY(0.98);
|
| 192 |
gStyle->SetTitleW(0.8);
|
| 193 |
gStyle->SetTitleH(0.06);
|
| 194 |
|
| 195 |
gStyle->SetErrorX(0);
|
| 196 |
gStyle->SetStatColor(0);
|
| 197 |
gStyle->SetStatBorderSize(0);
|
| 198 |
gStyle->SetStatX(0);
|
| 199 |
gStyle->SetStatY(0);
|
| 200 |
gStyle->SetStatW(0);
|
| 201 |
gStyle->SetStatH(0);
|
| 202 |
|
| 203 |
gStyle->SetTitleFont(22);
|
| 204 |
gStyle->SetLabelFont(22,"X");
|
| 205 |
gStyle->SetLabelFont(22,"Y");
|
| 206 |
gStyle->SetLabelFont(22,"Z");
|
| 207 |
gStyle->SetLabelSize(0.03,"X");
|
| 208 |
gStyle->SetLabelSize(0.03,"Y");
|
| 209 |
gStyle->SetLabelSize(0.03,"Z");
|
| 210 |
|
| 211 |
if (!backgd_ || !signal_)
|
| 212 |
{
|
| 213 |
std::cerr << "ERROR! &(background histogram)="<<backgd_
|
| 214 |
<< ", &(signal histogram)="<<signal_<<std::endl;
|
| 215 |
return;
|
| 216 |
}
|
| 217 |
|
| 218 |
//Draw MET for background, signal & data:
|
| 219 |
TCanvas * c1 = new TCanvas("","",600,600);
|
| 220 |
TPostScript * ps = new TPostScript( outputfilename_.c_str(),111);
|
| 221 |
|
| 222 |
//CLs vs x-section:
|
| 223 |
//int n_points = signals_.size();
|
| 224 |
int n_points = 51;
|
| 225 |
double min = 999999.;
|
| 226 |
double max = -99999.;
|
| 227 |
for (std::vector<double>::const_iterator it=ScanPar_.begin();it!=ScanPar_.end();++it){
|
| 228 |
if (*it<min) min = *it;
|
| 229 |
if (*it>max) max = *it;
|
| 230 |
}
|
| 231 |
--n_points;
|
| 232 |
|
| 233 |
TLine * line = new TLine( );
|
| 234 |
|
| 235 |
char * title = new char[256];
|
| 236 |
sprintf(title,";%s;CLs",ScanParName_.c_str());
|
| 237 |
TH1D * axis_cls = new TH1D("axis", title,0,min,max);
|
| 238 |
TGraph * cls_obs = new TGraph(1);
|
| 239 |
TGraph * cls_exp = new TGraph(1);
|
| 240 |
sprintf(title,";%s;-2 ln Q",ScanParName_.c_str());
|
| 241 |
TH1D * axis_lnQ = new TH1D("-2lnQ_obs", title,0,min,max);
|
| 242 |
TGraph * lnQ_obs = new TGraph(1);
|
| 243 |
TGraph * lnQ_exp_b = new TGraph(1);
|
| 244 |
TGraph * lnQ_exp_sb = new TGraph(1);
|
| 245 |
double y_cls_exp1[n_points*2+2];
|
| 246 |
double x_cls[n_points*2+2];
|
| 247 |
double y_cls_exp2[n_points*2+2];
|
| 248 |
double y_lnQ_exp1[n_points*2+2];
|
| 249 |
double y_lnQ_exp2[n_points*2+2];
|
| 250 |
double x_lnQ[n_points*2+2];
|
| 251 |
int i_cls=0; // i_lnQ=0;
|
| 252 |
|
| 253 |
|
| 254 |
TH1D * backgd = (TH1D*)backgd_->Clone();
|
| 255 |
TH1D * data = (TH1D*)data_->Clone();
|
| 256 |
data->SetMarkerStyle(8);
|
| 257 |
backgd->SetTitle("");
|
| 258 |
|
| 259 |
TTable limits("Limits of parameter scan","|");
|
| 260 |
limits.AddColumn<double>(ScanParName_);
|
| 261 |
limits.AddColumn<double>(" #s ");
|
| 262 |
limits.AddColumn<double>("+-ds");
|
| 263 |
limits.AddColumn<double>("CL_s ");limits.SetPrecision(3,5);
|
| 264 |
limits.AddColumn<double>("CL_sb");limits.SetPrecision(4,5);
|
| 265 |
limits.AddColumn<double>("CL_b ");limits.SetPrecision(5,5);
|
| 266 |
limits.AddColumn<double>("<CL_sb>_b");limits.SetPrecision(6,5);
|
| 267 |
limits.AddColumn<double>("<CL_b>_b");limits.SetPrecision(7,5);
|
| 268 |
limits.AddColumn<double>("<CL_s>_b-2s");limits.SetPrecision(8,5);
|
| 269 |
limits.AddColumn<double>("<CL_s>_b-1s");limits.SetPrecision(9,5);
|
| 270 |
limits.AddColumn<double>("<CL_s>_b");limits.SetPrecision(10,5);
|
| 271 |
limits.AddColumn<double>("<CL_s>_b+1s");limits.SetPrecision(11,5);
|
| 272 |
limits.AddColumn<double>("<CL_s>_b+2s");limits.SetPrecision(12,5);
|
| 273 |
limits.AddColumn<double>("-2lnQ");
|
| 274 |
limits.AddColumn<double>("<lnQ_b>");
|
| 275 |
limits.AddColumn<double>("<lnQ_sb>");
|
| 276 |
|
| 277 |
std::cout << "Scanning " << ScanParName_ << " in " << n_points << " steps from " << min << " to "<< max << ":"<<std::endl;
|
| 278 |
for (int i=0; i<=n_points; ++i){
|
| 279 |
|
| 280 |
double x;
|
| 281 |
TH1D * signal;
|
| 282 |
|
| 283 |
if (n_points-1==(int)signals_.size()){
|
| 284 |
x = ScanPar_[i];
|
| 285 |
signal = (TH1D*)signals_[i]->Clone();
|
| 286 |
} else {
|
| 287 |
x = min + (max-min)/(double)(n_points)*(double)i;
|
| 288 |
GetXandYbyInterpolation( x, signal);
|
| 289 |
}
|
| 290 |
|
| 291 |
if (!signal) { std::cout << "Error at x="<<x<<"; skipping..."<< std::endl; continue;}
|
| 292 |
|
| 293 |
signal->SetLineColor(2);
|
| 294 |
signal->SetLineWidth(3);
|
| 295 |
c1->SetLogy(0);
|
| 296 |
if (!i) backgd->GetXaxis()->SetTitle( signal->GetXaxis()->GetTitle() );
|
| 297 |
backgd->Draw("h");
|
| 298 |
signal->Draw("h,same");
|
| 299 |
data->Draw("pe,same");
|
| 300 |
c1->Update();
|
| 301 |
ps->NewPage();
|
| 302 |
|
| 303 |
//cout the CL's:
|
| 304 |
TLimitDataSource* mydatasource = new TLimitDataSource(signal,backgd,data);
|
| 305 |
TConfidenceLevel *myconfidence = TLimit::ComputeLimit(mydatasource,fNMC_,stat_);
|
| 306 |
|
| 307 |
double cls = myconfidence->CLs();
|
| 308 |
double cls_b = myconfidence->GetExpectedCLs_b();
|
| 309 |
double cls_b_p1 = myconfidence->GetExpectedCLs_b(1);
|
| 310 |
double cls_b_n1 = myconfidence->GetExpectedCLs_b(-1);
|
| 311 |
double cls_b_p2 = myconfidence->GetExpectedCLs_b(2);
|
| 312 |
double cls_b_n2 = myconfidence->GetExpectedCLs_b(-2);
|
| 313 |
Sort(cls_b_n1, cls_b_p1);
|
| 314 |
Sort(cls_b_n2, cls_b_p2);
|
| 315 |
|
| 316 |
double ds=0, db=0;
|
| 317 |
for (int a=0; a<signal->GetNbinsX()+1; ++a) ds+=signal->GetBinError(a);
|
| 318 |
for (int a=0; a<backgd->GetNbinsX()+1; ++a) db+=backgd->GetBinError(a);
|
| 319 |
|
| 320 |
limits.Fill(x,
|
| 321 |
signal->Integral(),ds,
|
| 322 |
backgd->Integral(),db,
|
| 323 |
cls,myconfidence->CLsb(),myconfidence->CLb(),
|
| 324 |
myconfidence->GetExpectedCLsb_b(),myconfidence->GetExpectedCLb_b(),
|
| 325 |
//myconfidence->GetExpectedStatistic_b(1),myconfidence->GetExpectedStatistic_b(-1),myconfidence->GetExpectedStatistic_b(2),myconfidence->GetExpectedStatistic_b(-2),
|
| 326 |
cls_b_n2,cls_b_n1,cls_b,cls_b_p1,cls_b_p2,
|
| 327 |
myconfidence->GetStatistic(),myconfidence->GetExpectedStatistic_b(),myconfidence->GetExpectedStatistic_sb()
|
| 328 |
);
|
| 329 |
|
| 330 |
//Draw the expected and observed test statistics:
|
| 331 |
c1->SetLogy(1);
|
| 332 |
TH1D h("TConfidenceLevel_Draw","",50,0,0);
|
| 333 |
double * fTSB = myconfidence->GetTestStatistic_B();
|
| 334 |
double * fTSS = myconfidence->GetTestStatistic_SB();
|
| 335 |
for (int j=0; j<fNMC_; j++) {
|
| 336 |
h.Fill(-2*(fTSB[j]-myconfidence->GetStot() ));
|
| 337 |
h.Fill(-2*(fTSS[j]-myconfidence->GetStot() ));
|
| 338 |
}
|
| 339 |
sprintf(title,"b_hist_%d",plotindex_++);
|
| 340 |
TH1D* b_hist = new TH1D(title, ";-2 ln Q;normalized p.d.f.",50,h.GetXaxis()->GetXmin(),h.GetXaxis()->GetXmax());
|
| 341 |
sprintf(title,"sb_hist_%d",plotindex_++);
|
| 342 |
TH1D* sb_hist = new TH1D(title,";-2 ln Q;normalized p.d.f.",50,h.GetXaxis()->GetXmin(),h.GetXaxis()->GetXmax());
|
| 343 |
for (int j=0; j<fNMC_; j++) {
|
| 344 |
b_hist->Fill(-2*(fTSB[j]-myconfidence->GetStot() ));
|
| 345 |
sb_hist->Fill(-2*(fTSS[j]-myconfidence->GetStot() ));
|
| 346 |
}
|
| 347 |
b_hist->Scale(1.0/b_hist->Integral());
|
| 348 |
sb_hist->Scale(1.0/sb_hist->Integral());
|
| 349 |
b_hist->GetYaxis()->SetTitleOffset(1.3);
|
| 350 |
b_hist->SetMinimum(0.001);
|
| 351 |
b_hist->SetMaximum(3*b_hist->GetMaximum());
|
| 352 |
b_hist->SetLineWidth(3);
|
| 353 |
sb_hist->SetLineWidth(3);
|
| 354 |
b_hist->SetLineColor(kBlue);
|
| 355 |
sb_hist->SetLineColor( 28 );
|
| 356 |
line->SetLineWidth(3);
|
| 357 |
TLegend * leg = new TLegend(0.11,0.78,0.48,0.89);
|
| 358 |
leg->SetBorderSize(0);
|
| 359 |
leg->SetFillColor(0);
|
| 360 |
leg->AddEntry(line,"Observed ","l");
|
| 361 |
leg->AddEntry(b_hist,"Expected background-only ","l");
|
| 362 |
leg->AddEntry(sb_hist,"Expected signal+background ","l");
|
| 363 |
b_hist->Draw("h");
|
| 364 |
line->DrawLine(myconfidence->GetStatistic(),b_hist->GetMinimum(),
|
| 365 |
myconfidence->GetStatistic(),0.6*b_hist->GetMaximum());
|
| 366 |
leg->Draw("same");
|
| 367 |
sb_hist->Draw("h,same");
|
| 368 |
c1->Update();
|
| 369 |
ps->NewPage();
|
| 370 |
|
| 371 |
//Fill Test-Statistics hists
|
| 372 |
lnQ_obs->SetPoint( i, x, myconfidence->GetStatistic() );
|
| 373 |
lnQ_exp_b->SetPoint( i, x, myconfidence->GetExpectedStatistic_b() );
|
| 374 |
lnQ_exp_sb->SetPoint( i, x, myconfidence->GetExpectedStatistic_sb() );
|
| 375 |
y_lnQ_exp1[i] = myconfidence->GetExpectedStatistic_b(1);
|
| 376 |
y_lnQ_exp1[2*n_points-i+1] = myconfidence->GetExpectedStatistic_b(-1);
|
| 377 |
y_lnQ_exp2[i] = myconfidence->GetExpectedStatistic_b(2);
|
| 378 |
y_lnQ_exp2[2*n_points-i+1] = myconfidence->GetExpectedStatistic_b(-2);
|
| 379 |
x_lnQ[i] = x;
|
| 380 |
x_lnQ[2*n_points-i+1] = x;
|
| 381 |
|
| 382 |
//Fill CLs hists:
|
| 383 |
if (cls<=1. && cls_b<=1. && cls_b_p1<999. &&cls_b_n1<999. &&cls_b_p2<999. &&cls_b_n2<999. &&
|
| 384 |
cls_b>0.) {
|
| 385 |
cls_obs->SetPoint( i_cls, x, cls );
|
| 386 |
cls_exp->SetPoint( i_cls, x, cls_b );
|
| 387 |
y_cls_exp1[i_cls] = cls_b_p1;
|
| 388 |
y_cls_exp1[2*n_points-i_cls+1] =cls_b_n1;
|
| 389 |
y_cls_exp2[i_cls] = cls_b_p2;
|
| 390 |
y_cls_exp2[2*n_points-i_cls+1] = cls_b_n2;
|
| 391 |
x_cls[i_cls] = x;
|
| 392 |
x_cls[2*n_points-i_cls+1] = x;
|
| 393 |
|
| 394 |
++i_cls;
|
| 395 |
}
|
| 396 |
|
| 397 |
delete myconfidence;
|
| 398 |
delete mydatasource;
|
| 399 |
delete signal;
|
| 400 |
}
|
| 401 |
delete data;
|
| 402 |
delete backgd;
|
| 403 |
|
| 404 |
std::cout << limits << std::endl;
|
| 405 |
|
| 406 |
//-2 ln Q vs mass:
|
| 407 |
//c1->SetLogx(1);
|
| 408 |
c1->SetLogy(0);
|
| 409 |
TGraph * lnQ_exp1 = new TGraph(2*n_points+2,x_lnQ,y_lnQ_exp1);
|
| 410 |
TGraph * lnQ_exp2 = new TGraph(2*n_points+2,x_lnQ,y_lnQ_exp2);
|
| 411 |
lnQ_exp2->SetLineColor(5);
|
| 412 |
lnQ_exp2->SetFillColor(5);
|
| 413 |
lnQ_exp1->SetLineColor(3);
|
| 414 |
lnQ_exp1->SetFillColor(3);
|
| 415 |
lnQ_obs->SetLineColor(kRed);
|
| 416 |
lnQ_obs->SetLineWidth(3);
|
| 417 |
lnQ_exp_b->SetLineColor(kBlue);
|
| 418 |
lnQ_exp_b->SetLineStyle(3);
|
| 419 |
lnQ_exp_b->SetLineWidth(3);
|
| 420 |
lnQ_exp_sb->SetLineColor( 28 );
|
| 421 |
lnQ_exp_sb->SetLineStyle(3);
|
| 422 |
lnQ_exp_sb->SetLineWidth(3);
|
| 423 |
axis_lnQ->SetMaximum( lnQ_exp2->GetYaxis()->GetXmax() );
|
| 424 |
axis_lnQ->SetMinimum( lnQ_exp_sb->GetYaxis()->GetXmin() );
|
| 425 |
line->SetLineWidth(1);
|
| 426 |
TLegend * lgd = new TLegend(0.5,0.28,0.89,0.4);
|
| 427 |
lgd->SetBorderSize(0);
|
| 428 |
lgd->SetFillColor(0);
|
| 429 |
if (isPseudoData) lgd->AddEntry(lnQ_obs,"Observed (pseudo data)","l");
|
| 430 |
else lgd->AddEntry(lnQ_obs,"Observed","l");
|
| 431 |
lgd->AddEntry(lnQ_exp_b,"Expected background-only","l");
|
| 432 |
lgd->AddEntry(lnQ_exp_sb,"Expected signal+background","l");
|
| 433 |
axis_lnQ->Draw("");
|
| 434 |
lnQ_exp2->Draw("lf,same");
|
| 435 |
lnQ_exp1->Draw("lf,same");
|
| 436 |
lnQ_obs->Draw("l,same");
|
| 437 |
line->DrawLine(min,0.0,max,0.0);
|
| 438 |
lnQ_exp_b->Draw("l,same");
|
| 439 |
lnQ_exp_sb->Draw("l,same");
|
| 440 |
lgd->Draw("same");
|
| 441 |
if (doeps) c1->SaveAs("pictures/lnQ_vs_param.eps");
|
| 442 |
c1->Update();
|
| 443 |
ps->NewPage();
|
| 444 |
|
| 445 |
//c1->SetLogx(1);
|
| 446 |
double exp_xsec_limit=min;
|
| 447 |
double obs_xsec_limit=min;
|
| 448 |
std::cout<<"Limits on "<<ScanParName_<<":"<<std::endl;
|
| 449 |
exp_xsec_limit = GetExpectedXsecLimit(0.05, min, max);
|
| 450 |
std::cout<<"Expected: "<<exp_xsec_limit<<" @ 95%CL in " << ScanParName_ << std::endl;
|
| 451 |
obs_xsec_limit = GetObservedXsecLimit(0.05, min, max);
|
| 452 |
std::cout<<"Observed: "<<obs_xsec_limit<<" @ 95%CL in " << ScanParName_ << std::endl;
|
| 453 |
|
| 454 |
//double par[] = {1.6449};//<->5%
|
| 455 |
//TSignificance eSignif(backgd->Integral(), 0);
|
| 456 |
//double exp_signif = TSolve<TSignificance>( &eSignif, &TSignificance::operator(), par, 0.1, 10.)();
|
| 457 |
////double obs_signif = exp_signif/(data->Integral()-backgd->Integral());
|
| 458 |
//std::cout<<"Expected s: "<< exp_signif/signal->Integral() <<" * signal cross section @ 95%CL significance" << std::endl;
|
| 459 |
////std::cout<<"Observed s: "<< obs_signif <<" * signal cross section @ 95%CL" << std::endl;
|
| 460 |
|
| 461 |
c1->SetLogy(1);
|
| 462 |
//remove empty array elements, to get a smooth TGraph:
|
| 463 |
--i_cls;
|
| 464 |
double ny_cls_exp1[i_cls*2];
|
| 465 |
double nx_cls[i_cls*2];
|
| 466 |
double ny_cls_exp2[i_cls*2];
|
| 467 |
for (int i=0; i<=i_cls; ++i) {
|
| 468 |
ny_cls_exp1[i] = y_cls_exp1[i];
|
| 469 |
ny_cls_exp2[i] = y_cls_exp2[i];
|
| 470 |
nx_cls[i] = x_cls[i];
|
| 471 |
ny_cls_exp1[i_cls+i] = y_cls_exp1[2*n_points-i_cls+i];
|
| 472 |
ny_cls_exp2[i_cls+i] = y_cls_exp2[2*n_points-i_cls+i];
|
| 473 |
nx_cls[i_cls+i] = x_cls[2*n_points-i_cls+i];
|
| 474 |
}
|
| 475 |
|
| 476 |
TGraph * cls_exp1 = new TGraph(2*i_cls,nx_cls,ny_cls_exp1);
|
| 477 |
TGraph * cls_exp2 = new TGraph(2*i_cls,nx_cls,ny_cls_exp2);
|
| 478 |
cls_exp2->SetLineColor(5);
|
| 479 |
cls_exp2->SetFillColor(5);
|
| 480 |
cls_exp1->SetLineColor(3);
|
| 481 |
cls_exp1->SetFillColor(3);
|
| 482 |
cls_obs->SetLineColor(kRed);
|
| 483 |
cls_exp->SetLineColor(kBlue);
|
| 484 |
cls_exp->SetLineStyle(3);
|
| 485 |
cls_exp->SetLineWidth(3);
|
| 486 |
cls_obs->SetLineWidth(3);
|
| 487 |
//axis_cls->SetMinimum( cls_exp2->GetYaxis()->GetXmin() );
|
| 488 |
axis_cls->SetMinimum( 0.000001 );
|
| 489 |
axis_cls->SetMaximum(1.1);
|
| 490 |
TLegend * lg = new TLegend(0.5,0.28,0.89,0.4);
|
| 491 |
lg->SetBorderSize(0);
|
| 492 |
lg->SetFillColor(0);
|
| 493 |
if (isPseudoData) lg->AddEntry(cls_obs,"Observed (pseudo data)","l");
|
| 494 |
else lg->AddEntry(cls_obs,"Observed","l");
|
| 495 |
lg->AddEntry(cls_exp,"Expected background-only ","l");
|
| 496 |
axis_cls->Draw("l");
|
| 497 |
cls_exp2->Draw("lf,same");
|
| 498 |
cls_exp1->Draw("lf,same");
|
| 499 |
cls_exp->Draw("l,same");
|
| 500 |
cls_obs->Draw("l,same");
|
| 501 |
lg->Draw("same");
|
| 502 |
line->SetLineWidth(1);
|
| 503 |
line->DrawLine(min,0.05,max,0.05);
|
| 504 |
TArrow * arrow = new TArrow();
|
| 505 |
arrow->SetLineWidth(1);
|
| 506 |
if (obs_xsec_limit>min && obs_xsec_limit<max)
|
| 507 |
arrow->DrawArrow(obs_xsec_limit,0.05,obs_xsec_limit,axis_cls->GetMinimum(),0.03,">");
|
| 508 |
arrow->SetLineStyle(7);
|
| 509 |
if (exp_xsec_limit>min && exp_xsec_limit<max)
|
| 510 |
arrow->DrawArrow(exp_xsec_limit,50.*axis_cls->GetMinimum(),
|
| 511 |
exp_xsec_limit, axis_cls->GetMinimum(),0.03,">");
|
| 512 |
|
| 513 |
if (doeps) c1->SaveAs("pictures/CLs_vs_param.eps");
|
| 514 |
c1->Update();
|
| 515 |
ps->Close();
|
| 516 |
}
|
| 517 |
|
| 518 |
void cls::DrawVsXsec(bool doeps)
|
| 519 |
{
|
| 520 |
gStyle->SetHistFillColor(0);
|
| 521 |
gStyle->SetPalette(1);
|
| 522 |
gStyle->SetCanvasColor(0);
|
| 523 |
gStyle->SetCanvasBorderMode(0);
|
| 524 |
gStyle->SetPadColor(0);
|
| 525 |
gStyle->SetPadBorderMode(0);
|
| 526 |
gStyle->SetFrameBorderMode(0);
|
| 527 |
|
| 528 |
gStyle->SetTitleFillColor(0);
|
| 529 |
gStyle->SetTitleBorderSize(0);
|
| 530 |
gStyle->SetTitleX(0.10);
|
| 531 |
gStyle->SetTitleY(0.98);
|
| 532 |
gStyle->SetTitleW(0.8);
|
| 533 |
gStyle->SetTitleH(0.06);
|
| 534 |
|
| 535 |
gStyle->SetErrorX(0);
|
| 536 |
gStyle->SetStatColor(0);
|
| 537 |
gStyle->SetStatBorderSize(0);
|
| 538 |
gStyle->SetStatX(0);
|
| 539 |
gStyle->SetStatY(0);
|
| 540 |
gStyle->SetStatW(0);
|
| 541 |
gStyle->SetStatH(0);
|
| 542 |
|
| 543 |
gStyle->SetTitleFont(22);
|
| 544 |
gStyle->SetLabelFont(22,"X");
|
| 545 |
gStyle->SetLabelFont(22,"Y");
|
| 546 |
gStyle->SetLabelFont(22,"Z");
|
| 547 |
gStyle->SetLabelSize(0.03,"X");
|
| 548 |
gStyle->SetLabelSize(0.03,"Y");
|
| 549 |
gStyle->SetLabelSize(0.03,"Z");
|
| 550 |
|
| 551 |
if (!backgd_ || !signal_)
|
| 552 |
{
|
| 553 |
std::cerr << "ERROR! &(background histogram)="<<backgd_
|
| 554 |
<< ", &(signal histogram)="<<signal_<<std::endl;
|
| 555 |
return;
|
| 556 |
}
|
| 557 |
|
| 558 |
TH1D * backgd = (TH1D*)backgd_->Clone();
|
| 559 |
TH1D * signal = (TH1D*)signal_->Clone();
|
| 560 |
TH1D * data = (TH1D*)data_->Clone();
|
| 561 |
backgd->GetXaxis()->SetTitle( signal->GetXaxis()->GetTitle() );
|
| 562 |
backgd->SetTitle("");
|
| 563 |
|
| 564 |
std::cout << "DATA = " << data->Integral(0, data->GetXaxis()->GetNbins()+1) <<std::endl;
|
| 565 |
|
| 566 |
//Draw MET for background, signal & data:
|
| 567 |
TCanvas * c1 = new TCanvas("","",600,600);
|
| 568 |
TPostScript * ps = new TPostScript( outputfilename_.c_str(),111);
|
| 569 |
TLine * line = new TLine();
|
| 570 |
signal->SetLineColor(2);
|
| 571 |
signal->SetLineWidth(3);
|
| 572 |
data->SetMarkerStyle(8);
|
| 573 |
backgd->SetMinimum(0);
|
| 574 |
backgd->Draw("he");
|
| 575 |
signal->Draw("he,same");
|
| 576 |
data->Draw("pe,same");
|
| 577 |
c1->Update();
|
| 578 |
ps->NewPage();
|
| 579 |
|
| 580 |
//CLs vs x-section:
|
| 581 |
int n_points = 10;
|
| 582 |
double min = 0.01;
|
| 583 |
double max = 1;
|
| 584 |
|
| 585 |
//min *= signalxsec;
|
| 586 |
//max *= signalxsec;
|
| 587 |
|
| 588 |
TH1D * cls_obs = new TH1D("cls_obs", ";relative cross section k [k * #sigma];CLs",n_points,min,max);
|
| 589 |
// TH1D * cls_obs = new TH1D("cls_obs", ";signal cross section #sigma;CLs",n_points,min,max);
|
| 590 |
TH1D * cls_exp = new TH1D("cls_exp", ";relative cross section k [k * #sigma];CLs",n_points,min,max);
|
| 591 |
TH1D * lnQ_obs = new TH1D("lnQ_obs", ";relative cross section k [k * #sigma];-2 ln Q",n_points,min,max);
|
| 592 |
TH1D * lnQ_exp_b = new TH1D("lnQ_exp_b", ";relative cross section k [k * #sigma];-2 ln Q",n_points,min,max);
|
| 593 |
TH1D * lnQ_exp_sb = new TH1D("lnQ_exp_sb", ";relative cross section k [k * #sigma];-2 ln Q",n_points,min,max);
|
| 594 |
double y_cls_exp1[n_points*2+2];
|
| 595 |
double x_cls_exp[n_points*2+2];
|
| 596 |
double x_lnQ_exp[n_points*2+2];
|
| 597 |
double y_cls_exp2[n_points*2+2];
|
| 598 |
double y_lnQ_exp1[n_points*2+2];
|
| 599 |
double y_lnQ_exp2[n_points*2+2];
|
| 600 |
|
| 601 |
std::cout<<"Scanning signal coss-section from x = "<<(int)(min*100)<<"% till "<<(int)(max*100)<<"%:"<<std::endl;
|
| 602 |
TTable limits("Limits of cross-section scan","|");
|
| 603 |
//limits.AddColumn<int>("i");
|
| 604 |
limits.AddColumn<double>("x/sigma");
|
| 605 |
limits.AddColumn<double>(" #s ");
|
| 606 |
limits.AddColumn<double>("+-ds");
|
| 607 |
limits.AddColumn<double>(" #b ");
|
| 608 |
limits.AddColumn<double>("+-db");
|
| 609 |
limits.AddColumn<double>("CL_s");
|
| 610 |
limits.AddColumn<double>("<CL_b>_b");limits.SetPrecision(6,5);
|
| 611 |
limits.AddColumn<double>("<CL_sb>_b");limits.SetPrecision(7,5);
|
| 612 |
limits.AddColumn<double>("<lnQ_b>");
|
| 613 |
limits.AddColumn<double>("<lnQ_sb>");
|
| 614 |
limits.AddColumn<double>("<CL_s-2s>_b");limits.SetPrecision(10,5);
|
| 615 |
limits.AddColumn<double>("<CL_s-1s>_b");limits.SetPrecision(11,5);
|
| 616 |
limits.AddColumn<double>("<CL_s>_b");limits.SetPrecision(12,5);
|
| 617 |
limits.AddColumn<double>("<CL_s+1s>_b");limits.SetPrecision(13,5);
|
| 618 |
limits.AddColumn<double>("<CL_s+2s>_b");limits.SetPrecision(14,5);
|
| 619 |
char * title = new char[100];
|
| 620 |
int i_cls=0;
|
| 621 |
|
| 622 |
for (int i=0; i<=n_points; ++i){
|
| 623 |
if ((int)(100*(float)i/n_points)%10==0)
|
| 624 |
std::cout << std::setw(2)<<100*i/n_points << "% done"<<std::endl;
|
| 625 |
|
| 626 |
|
| 627 |
TH1D * sig = (TH1D*)signal_->Clone();
|
| 628 |
double x = min + (max-min)*i/n_points;
|
| 629 |
sig->Scale( x );
|
| 630 |
|
| 631 |
// double x = min + (double)i/n_points*(max-min);
|
| 632 |
// sig->Scale( x/signalxsec );
|
| 633 |
// x *= signalxsec;
|
| 634 |
|
| 635 |
TLimitDataSource * source = new TLimitDataSource();
|
| 636 |
if (!syst_) source->AddChannel(sig,backgd,data);
|
| 637 |
else source->AddChannel(sig,backgd,data,(TH1D*)esup_,(TH1D*)esdn_,(TH1D*)ebup_,(TH1D*)ebdn_,names_);
|
| 638 |
TConfidenceLevel * conf = TLimit::ComputeLimit(source,fNMC_,stat_);
|
| 639 |
|
| 640 |
//Draw plot
|
| 641 |
c1->SetLogy(0);
|
| 642 |
sig->SetLineColor(2);
|
| 643 |
sig->SetLineWidth(3);
|
| 644 |
data->SetMarkerStyle(8);
|
| 645 |
backgd->SetMinimum(0);
|
| 646 |
if (sig->GetMaximum()>backgd->GetMaximum()) backgd->SetMaximum(sig->GetMaximum()*sqrt(sig->GetMaximum()));
|
| 647 |
backgd->Draw("he");
|
| 648 |
sig->Draw("he,same");
|
| 649 |
data->Draw("pe,same");
|
| 650 |
if (doeps) {
|
| 651 |
sprintf(title,"pictures/N_for_x%.3f.eps",x);
|
| 652 |
c1->SaveAs(title);
|
| 653 |
}
|
| 654 |
c1->Update();
|
| 655 |
ps->NewPage();
|
| 656 |
|
| 657 |
//Draw the expected and observed test statistics:
|
| 658 |
c1->SetLogy(1);
|
| 659 |
sprintf(title,"xTConfidenceLevel_Draw_%i",plotindex_++);
|
| 660 |
TH1D h(title,"",50,0,0);
|
| 661 |
double * fTSB = conf->GetTestStatistic_B();
|
| 662 |
double * fTSS = conf->GetTestStatistic_SB();
|
| 663 |
for (int j=0; j<fNMC_; j++) {
|
| 664 |
h.Fill(-2*(fTSB[j]-conf->GetStot() ));
|
| 665 |
h.Fill(-2*(fTSS[j]-conf->GetStot() ));
|
| 666 |
}
|
| 667 |
sprintf(title,"xb_hist_%i",plotindex_++);
|
| 668 |
TH1D* b_hist = new TH1D(title, ";-2 ln Q;normalized p.d.f.",50,h.GetXaxis()->GetXmin(),h.GetXaxis()->GetXmax());
|
| 669 |
sprintf(title,"xsb_hist_%i",plotindex_++);
|
| 670 |
TH1D* sb_hist = new TH1D(title,";-2 ln Q;normalized p.d.f.",50,h.GetXaxis()->GetXmin(),h.GetXaxis()->GetXmax());
|
| 671 |
for (int j=0; j<fNMC_; j++) {
|
| 672 |
b_hist->Fill(-2*(fTSB[j]-conf->GetStot() ));
|
| 673 |
sb_hist->Fill(-2*(fTSS[j]-conf->GetStot() ));
|
| 674 |
}
|
| 675 |
b_hist->Scale(1.0/b_hist->Integral());
|
| 676 |
sb_hist->Scale(1.0/sb_hist->Integral());
|
| 677 |
b_hist->GetYaxis()->SetTitleOffset(1.3);
|
| 678 |
b_hist->SetMinimum( 1./fNMC_ );
|
| 679 |
b_hist->SetMaximum(3*b_hist->GetMaximum());
|
| 680 |
b_hist->SetLineWidth(3);
|
| 681 |
sb_hist->SetLineWidth(3);
|
| 682 |
b_hist->SetLineColor(kBlue);
|
| 683 |
sb_hist->SetLineColor( 28 );
|
| 684 |
line->DrawLine(conf->GetStatistic(),b_hist->GetMinimum(),
|
| 685 |
conf->GetStatistic(),0.6*b_hist->GetMaximum() );
|
| 686 |
line->SetLineWidth(3);
|
| 687 |
TLegend leg(0.11,0.78,0.48,0.89);
|
| 688 |
leg.SetBorderSize(0);
|
| 689 |
leg.SetFillColor(0);
|
| 690 |
leg.AddEntry(line,"Observed ","l");
|
| 691 |
leg.AddEntry(b_hist,"Expected background-only ","l");
|
| 692 |
leg.AddEntry(sb_hist,"Expected signal+background ","l");
|
| 693 |
sprintf(title,"#sigma_{s} #times %f",x);
|
| 694 |
|
| 695 |
b_hist->Draw("h");
|
| 696 |
line->Draw();
|
| 697 |
leg.Draw("same");
|
| 698 |
sb_hist->Draw("h,same");
|
| 699 |
if (doeps) {
|
| 700 |
sprintf(title,"pictures/lnQ_for_x%.3f.eps",x);
|
| 701 |
c1->SaveAs(title);
|
| 702 |
}
|
| 703 |
c1->Update();
|
| 704 |
ps->NewPage();
|
| 705 |
|
| 706 |
//Plot # of signal and backgd events of pseudo experiments:
|
| 707 |
c1->SetLogy(0);
|
| 708 |
double * psig = conf->GetSig();
|
| 709 |
double * pbgd = conf->GetBgd();
|
| 710 |
double ds=0, db=0;
|
| 711 |
for (int a=0; a<sig->GetNbinsX()+1; ++a) ds+=sig->GetBinError(a);
|
| 712 |
for (int a=0; a<backgd->GetNbinsX()+1; ++a) db+=backgd->GetBinError(a);
|
| 713 |
|
| 714 |
double pmin=999999, pmax=-999999;
|
| 715 |
for (int it=0; it<fNMC_; ++it){
|
| 716 |
if (psig[it]<pmin) pmin = psig[it];
|
| 717 |
if (psig[it]>pmax && psig[it]<99999.) pmax = psig[it];
|
| 718 |
}
|
| 719 |
for (int it=0; it<fNMC_; ++it){
|
| 720 |
if (pbgd[it]<pmin) pmin = pbgd[it];
|
| 721 |
if (pbgd[it]>pmax && pbgd[it]<99999.) pmax = pbgd[it];
|
| 722 |
}
|
| 723 |
sprintf(title,"pseudosig_%i",plotindex_++);
|
| 724 |
TH1F * pseudosig = new TH1F(title,";# events;# pseudo experiments",100,pmin-5,pmax+5);
|
| 725 |
sprintf(title,"pseudobgd_%i",plotindex_++);
|
| 726 |
TH1F * pseudobgd = new TH1F(title,";# events;# pseudo experiments",100,pmin-5,pmax+5);
|
| 727 |
for (int it=0; it<fNMC_; ++it){
|
| 728 |
pseudosig->Fill(psig[it]);
|
| 729 |
}
|
| 730 |
for (int it=0; it<fNMC_; ++it){
|
| 731 |
pseudobgd->Fill(pbgd[it]);
|
| 732 |
}
|
| 733 |
pseudosig->SetLineColor(2);
|
| 734 |
pseudosig->SetLineWidth(3);
|
| 735 |
pseudobgd->SetLineWidth(3);
|
| 736 |
pseudosig->Draw("h");
|
| 737 |
pseudobgd->Draw("h,same");
|
| 738 |
TLegend pleg(0.31,0.8,0.68,0.89);
|
| 739 |
pleg.SetBorderSize(0);
|
| 740 |
pleg.SetFillColor(0);
|
| 741 |
sprintf(title,"Signal %.2f#pm%.2f",sig->Integral(),ds);
|
| 742 |
pleg.AddEntry(pseudosig,title,"l");
|
| 743 |
sprintf(title,"Background %.2f#pm%.2f",backgd->Integral(),db);
|
| 744 |
pleg.AddEntry(pseudobgd,title,"l");
|
| 745 |
pleg.Draw("same");
|
| 746 |
if (doeps) {
|
| 747 |
sprintf(title,"pictures/pseudoN_for_x%.3f.eps",x);
|
| 748 |
c1->SaveAs(title);
|
| 749 |
}
|
| 750 |
c1->Update();
|
| 751 |
ps->NewPage();
|
| 752 |
delete pseudosig;
|
| 753 |
delete pseudobgd;
|
| 754 |
|
| 755 |
//Calculate CLs and lnQ:
|
| 756 |
double cls = conf->CLs();
|
| 757 |
double cls_b = conf->GetExpectedCLs_b();
|
| 758 |
double cls_b_p1 = conf->GetExpectedCLs_b(1);
|
| 759 |
double cls_b_n1 = conf->GetExpectedCLs_b(-1);
|
| 760 |
double cls_b_p2 = conf->GetExpectedCLs_b(2);
|
| 761 |
double cls_b_n2 = conf->GetExpectedCLs_b(-2);
|
| 762 |
Sort(cls_b_n1, cls_b_p1);
|
| 763 |
Sort(cls_b_n2, cls_b_p2);
|
| 764 |
if (cls_b>0 && cls_b<=1.) {
|
| 765 |
cls_obs->SetBinContent(i_cls, cls );
|
| 766 |
cls_exp->SetBinContent(i_cls, cls_b );
|
| 767 |
y_cls_exp1[i_cls] = cls_b_p1;
|
| 768 |
y_cls_exp1[n_points*2-i_cls+1] = cls_b_n1;
|
| 769 |
y_cls_exp2[i_cls] = cls_b_p2;
|
| 770 |
y_cls_exp2[n_points*2-i_cls+1] = cls_b_n2;
|
| 771 |
x_cls_exp[i_cls] = x;
|
| 772 |
x_cls_exp[n_points*2-i_cls+1] = x;
|
| 773 |
++i_cls;
|
| 774 |
}
|
| 775 |
|
| 776 |
lnQ_obs->SetBinContent( i, conf->GetStatistic() );
|
| 777 |
lnQ_exp_b->SetBinContent( i, conf->GetExpectedStatistic_b() );
|
| 778 |
lnQ_exp_sb->SetBinContent( i, conf->GetExpectedStatistic_sb() );
|
| 779 |
y_lnQ_exp1[i] = conf->GetExpectedStatistic_b(1);
|
| 780 |
y_lnQ_exp1[n_points*2-i+1] = conf->GetExpectedStatistic_b(-1);
|
| 781 |
y_lnQ_exp2[i] = conf->GetExpectedStatistic_b(2);
|
| 782 |
y_lnQ_exp2[n_points*2-i+1] = conf->GetExpectedStatistic_b(-2);
|
| 783 |
x_lnQ_exp[i] = x;
|
| 784 |
x_lnQ_exp[n_points*2-i+1] = x;
|
| 785 |
|
| 786 |
limits.Fill(x,
|
| 787 |
sig->Integral(),ds,
|
| 788 |
backgd->Integral(),db,
|
| 789 |
cls,
|
| 790 |
conf->GetExpectedCLb_b(),conf->GetExpectedCLsb_b(),
|
| 791 |
conf->GetExpectedStatistic_b(),conf->GetExpectedStatistic_sb(),
|
| 792 |
cls_b_n2,cls_b_n1,
|
| 793 |
cls_b,
|
| 794 |
cls_b_p1,cls_b_p2
|
| 795 |
);
|
| 796 |
delete conf;
|
| 797 |
delete source;
|
| 798 |
delete sig;
|
| 799 |
}
|
| 800 |
std::cout << limits << std::endl;
|
| 801 |
|
| 802 |
//-2 ln Q vs mass:
|
| 803 |
//c1->SetLogx(1);
|
| 804 |
c1->SetLogy(0);
|
| 805 |
TGraph * lnQ_exp1 = new TGraph(2*n_points,x_lnQ_exp,y_lnQ_exp1);
|
| 806 |
TGraph * lnQ_exp2 = new TGraph(2*n_points,x_lnQ_exp,y_lnQ_exp2);
|
| 807 |
lnQ_exp2->SetLineColor(5);
|
| 808 |
lnQ_exp2->SetFillColor(5);
|
| 809 |
lnQ_exp1->SetLineColor(3);
|
| 810 |
lnQ_exp1->SetFillColor(3);
|
| 811 |
lnQ_obs->SetLineColor(kRed);
|
| 812 |
lnQ_obs->SetLineWidth(3);
|
| 813 |
lnQ_exp_b->SetLineColor(kBlue);
|
| 814 |
lnQ_exp_b->SetLineStyle(3);
|
| 815 |
lnQ_exp_b->SetLineWidth(3);
|
| 816 |
lnQ_exp_sb->SetLineColor( 28 );
|
| 817 |
lnQ_exp_sb->SetLineStyle(3);
|
| 818 |
lnQ_exp_sb->SetLineWidth(3);
|
| 819 |
lnQ_obs->SetMinimum(-20);
|
| 820 |
line->SetLineWidth(1);
|
| 821 |
TLegend * lgd = new TLegend(0.5,0.28,0.89,0.4);
|
| 822 |
lgd->SetBorderSize(0);
|
| 823 |
lgd->SetFillColor(0);
|
| 824 |
if (isPseudoData) lgd->AddEntry(lnQ_obs,"Observed (pseudo data)","l");
|
| 825 |
else lgd->AddEntry(lnQ_obs,"Observed","l");
|
| 826 |
lgd->AddEntry(lnQ_exp_b,"Expected background-only","l");
|
| 827 |
lgd->AddEntry(lnQ_exp_sb,"Expected signal+background","l");
|
| 828 |
lnQ_obs->Draw("c");
|
| 829 |
lnQ_exp2->Draw("lf,same");
|
| 830 |
lnQ_exp1->Draw("lf,same");
|
| 831 |
lnQ_obs->Draw("c,same");
|
| 832 |
line->DrawLine(100,0.0,500,0.0);
|
| 833 |
lnQ_exp_b->Draw("c,same");
|
| 834 |
lnQ_exp_sb->Draw("c,same");
|
| 835 |
lgd->Draw("same");
|
| 836 |
//if (doeps)
|
| 837 |
c1->SaveAs("pictures/lnQvsxsec.eps");
|
| 838 |
c1->Update();
|
| 839 |
ps->NewPage();
|
| 840 |
|
| 841 |
//c1->SetLogx(1);
|
| 842 |
double exp_xsec_limit = 0;
|
| 843 |
double obs_xsec_limit = 0;
|
| 844 |
std::cout<<"Cross-section limit:"<<std::endl;
|
| 845 |
//exp_xsec_limit = GetExpectedXsecLimit(0.05, min, max);
|
| 846 |
//obs_xsec_limit = GetObservedXsecLimit(0.05, min, max);
|
| 847 |
std::cout<<"Expected: "<<exp_xsec_limit<<" * signal cross section @ 95%CL" << std::endl;
|
| 848 |
std::cout<<"Observed: "<<obs_xsec_limit<<" * signal cross section @ 95%CL" << std::endl;
|
| 849 |
|
| 850 |
//double par[] = {1.6449};//<->5%
|
| 851 |
//TSignificance eSignif(backgd->Integral(), 0);
|
| 852 |
//double exp_signif = TSolve<TSignificance>( &eSignif, &TSignificance::operator(), par, 0.1, 10.)();
|
| 853 |
////double obs_signif = exp_signif/(data->Integral()-backgd->Integral());
|
| 854 |
//std::cout<<"Expected s: "<< exp_signif/signal->Integral() <<" * signal cross section @ 95%CL significance" << std::endl;
|
| 855 |
////std::cout<<"Observed s: "<< obs_signif <<" * signal cross section @ 95%CL" << std::endl;
|
| 856 |
//remove empty array elements, to get a smooth TGraph:
|
| 857 |
|
| 858 |
--i_cls;
|
| 859 |
double ny_cls_exp1[i_cls*2];
|
| 860 |
double nx_cls[i_cls*2];
|
| 861 |
double ny_cls_exp2[i_cls*2];
|
| 862 |
for (int i=0; i<=i_cls; ++i) {
|
| 863 |
ny_cls_exp1[i] = y_cls_exp1[i];
|
| 864 |
ny_cls_exp2[i] = y_cls_exp2[i];
|
| 865 |
nx_cls[i] = x_cls_exp[i];
|
| 866 |
ny_cls_exp1[i_cls+i] = y_cls_exp1[2*n_points-i_cls+i];
|
| 867 |
ny_cls_exp2[i_cls+i] = y_cls_exp2[2*n_points-i_cls+i];
|
| 868 |
nx_cls[i_cls+i] = x_cls_exp[2*n_points-i_cls+i];
|
| 869 |
}
|
| 870 |
|
| 871 |
c1->SetLogy(1);
|
| 872 |
TGraph * cls_exp1 = new TGraph(2*i_cls,nx_cls,ny_cls_exp1);
|
| 873 |
TGraph * cls_exp2 = new TGraph(2*i_cls,nx_cls,ny_cls_exp2);
|
| 874 |
cls_exp2->SetLineColor(5);
|
| 875 |
cls_exp2->SetFillColor(5);
|
| 876 |
cls_exp1->SetLineColor(3);
|
| 877 |
cls_exp1->SetFillColor(3);
|
| 878 |
cls_obs->SetLineColor(kRed);
|
| 879 |
cls_exp->SetLineColor(kBlue);
|
| 880 |
cls_exp->SetLineStyle(3);
|
| 881 |
cls_exp->SetLineWidth(3);
|
| 882 |
cls_obs->SetLineWidth(3);
|
| 883 |
//cls_obs->SetMinimum(0.00001);
|
| 884 |
cls_exp->SetMaximum(1.1);
|
| 885 |
TLegend * lg = new TLegend(0.5,0.28,0.89,0.4);
|
| 886 |
lg->SetBorderSize(0);
|
| 887 |
lg->SetFillColor(0);
|
| 888 |
if (isPseudoData) lg->AddEntry(cls_obs,"Observed (pseudo data)","l");
|
| 889 |
else lg->AddEntry(cls_obs,"Observed","l");
|
| 890 |
lg->AddEntry(cls_exp,"Expected background-only ","l");
|
| 891 |
cls_exp->Draw("c");
|
| 892 |
cls_exp2->Draw("lf,same");
|
| 893 |
cls_exp1->Draw("lf,same");
|
| 894 |
cls_obs->Draw("c,same");
|
| 895 |
cls_exp->Draw("c,same");
|
| 896 |
lg->Draw("same");
|
| 897 |
line->SetLineWidth(1);
|
| 898 |
line->DrawLine(min,0.05,max,0.05);
|
| 899 |
TArrow * arrow = new TArrow();
|
| 900 |
arrow->SetLineWidth(1);
|
| 901 |
if (obs_xsec_limit) arrow->DrawArrow(obs_xsec_limit,0.05,obs_xsec_limit,cls_obs->GetYaxis()->GetXmin(),0.03,">");
|
| 902 |
arrow->SetLineStyle(7);
|
| 903 |
if (exp_xsec_limit) arrow->DrawArrow(exp_xsec_limit,50.*cls_obs->GetMinimum(),
|
| 904 |
exp_xsec_limit, cls_obs->GetYaxis()->GetXmin(),0.03,">");
|
| 905 |
|
| 906 |
//if (doeps)
|
| 907 |
c1->SaveAs("pictures/CLsvsxsec.eps");
|
| 908 |
c1->Update();
|
| 909 |
delete data;
|
| 910 |
delete backgd;
|
| 911 |
delete signal;
|
| 912 |
|
| 913 |
ps->Close();
|
| 914 |
|
| 915 |
}
|
| 916 |
|
| 917 |
double cls::GetTotalStatError(const TH1* arg)
|
| 918 |
{
|
| 919 |
if (!arg) return 0;
|
| 920 |
TH1F * h = (TH1F*)arg->Clone();
|
| 921 |
h->Rebin( h->GetXaxis()->GetNbins() );
|
| 922 |
double result = sqrt(pow(h->GetBinError(0),2)+pow(h->GetBinError(1),2)+pow(h->GetBinError(2),2));
|
| 923 |
delete h;
|
| 924 |
return result;
|
| 925 |
}
|
| 926 |
|
| 927 |
void cls::WriteResult(ConfigFile * config)
|
| 928 |
{
|
| 929 |
double min=0.01; //min. limit on xsection in which CLs = 95% is searched for
|
| 930 |
double max=1000; //max. limit on xsection in which CLs = 95% is searched for
|
| 931 |
double xsec = config->read<double>("Xsection",1.0);
|
| 932 |
|
| 933 |
TH1D * backgd = (TH1D*)backgd_->Clone();
|
| 934 |
TH1D * signal = (TH1D*)signal_->Clone();
|
| 935 |
TH1D * data = (TH1D*)data_->Clone();
|
| 936 |
double nsignal = signal->Integral(0,signal->GetXaxis()->GetNbins()+1);
|
| 937 |
|
| 938 |
//std::cout << "...calculating ExpectedXsecLimit" <<std::endl;
|
| 939 |
double exp_xsec_limit = GetExpectedXsecLimit(0.05, min, max);
|
| 940 |
//std::cout << "...calculating ObservedXsecLimit" <<std::endl;
|
| 941 |
double obs_xsec_limit = GetObservedXsecLimit(0.05, min, max);
|
| 942 |
|
| 943 |
config->add("ExpXsecLimit", exp_xsec_limit * xsec);
|
| 944 |
config->add("ObsXsecLimit", obs_xsec_limit * xsec);
|
| 945 |
config->add("ExpNsigLimit", exp_xsec_limit * nsignal);
|
| 946 |
config->add("ObsNsigLimit", obs_xsec_limit * nsignal);
|
| 947 |
config->add("CLsUseStat", stat_);
|
| 948 |
|
| 949 |
if (obs_xsec_limit>=min && obs_xsec_limit<=max) {
|
| 950 |
TH1D * sig = (TH1D*)signal_->Clone();
|
| 951 |
sig->Scale( obs_xsec_limit );
|
| 952 |
TLimitDataSource * source = new TLimitDataSource();
|
| 953 |
if (!syst_) source->AddChannel(sig,backgd,data);
|
| 954 |
else source->AddChannel(sig,backgd,data,(TH1D*)esup_,(TH1D*)esdn_,(TH1D*)ebup_,(TH1D*)ebdn_,names_);
|
| 955 |
TConfidenceLevel * conf = TLimit::ComputeLimit(source,fNMC_,stat_);
|
| 956 |
config->add("CLs@obs", conf->CLs());
|
| 957 |
config->add("CLs_b@obs", conf->GetExpectedCLs_b());
|
| 958 |
config->add("CLs_b_p1@obs", conf->GetExpectedCLs_b(1));
|
| 959 |
config->add("CLs_b_n1@obs", conf->GetExpectedCLs_b(-1));
|
| 960 |
config->add("CLs_b_p2@obs", conf->GetExpectedCLs_b(2));
|
| 961 |
config->add("CLs_b_n2@obs", conf->GetExpectedCLs_b(-2));
|
| 962 |
config->add("CLb_b@obs", conf->GetExpectedCLb_b());
|
| 963 |
config->add("CLsb_b@obs", conf->GetExpectedCLsb_b());
|
| 964 |
config->add("-2lnQ_b@obs", conf->GetExpectedStatistic_b());
|
| 965 |
config->add("-2lnQ_sb@obs", conf->GetExpectedStatistic_sb());
|
| 966 |
delete sig;
|
| 967 |
delete conf;
|
| 968 |
delete source;
|
| 969 |
}
|
| 970 |
if (exp_xsec_limit>=min && exp_xsec_limit<=max) {
|
| 971 |
TH1D * sig = (TH1D*)signal_->Clone();
|
| 972 |
sig->Scale( exp_xsec_limit );
|
| 973 |
TLimitDataSource * source = new TLimitDataSource();
|
| 974 |
if (!syst_) source->AddChannel(sig,backgd,data);
|
| 975 |
else source->AddChannel(sig,backgd,data,(TH1D*)esup_,(TH1D*)esdn_,(TH1D*)ebup_,(TH1D*)ebdn_,names_);
|
| 976 |
TConfidenceLevel * conf = TLimit::ComputeLimit(source,fNMC_,stat_);
|
| 977 |
config->add("CLs@exp", conf->CLs());
|
| 978 |
config->add("CLs_b@exp", conf->GetExpectedCLs_b());
|
| 979 |
config->add("CLs_b_p1@exp", conf->GetExpectedCLs_b(1));
|
| 980 |
config->add("CLs_b_n1@exp", conf->GetExpectedCLs_b(-1));
|
| 981 |
config->add("CLs_b_p2@exp", conf->GetExpectedCLs_b(2));
|
| 982 |
config->add("CLs_b_n2@exp", conf->GetExpectedCLs_b(-2));
|
| 983 |
config->add("CLb_b@exp", conf->GetExpectedCLb_b());
|
| 984 |
config->add("CLsb_b@exp", conf->GetExpectedCLsb_b());
|
| 985 |
config->add("-2lnQ_b@exp", conf->GetExpectedStatistic_b());
|
| 986 |
config->add("-2lnQ_sb@exp", conf->GetExpectedStatistic_sb());
|
| 987 |
delete sig;
|
| 988 |
delete conf;
|
| 989 |
delete source;
|
| 990 |
}
|
| 991 |
delete backgd;
|
| 992 |
delete signal;
|
| 993 |
delete data;
|
| 994 |
}
|
| 995 |
|
| 996 |
void cls::WriteResult(const std::string out)
|
| 997 |
{
|
| 998 |
if (out=="") {
|
| 999 |
std::cerr<<"ERROR [cls::WriteResult(const std::string out="<<out<<")]: No filename and no signal parameters specified!" <<std::endl;
|
| 1000 |
return;
|
| 1001 |
}
|
| 1002 |
if (!backgd_ || !signal_) {
|
| 1003 |
std::cerr<<"ERROR [cls::WriteResult(const std::string out="<<out<<")]: No signal ("
|
| 1004 |
<<signal_<<") or no background ("<<backgd_<<") hypothesis specified!" <<std::endl;
|
| 1005 |
return;
|
| 1006 |
}
|
| 1007 |
std::stringstream ss;
|
| 1008 |
ss << out;
|
| 1009 |
ConfigFile * res = new ConfigFile();
|
| 1010 |
WriteResult( res );
|
| 1011 |
|
| 1012 |
//write stuff:
|
| 1013 |
time_t rawtime;
|
| 1014 |
struct tm * timeinfo;
|
| 1015 |
time ( &rawtime );
|
| 1016 |
timeinfo = localtime ( &rawtime );
|
| 1017 |
ofstream ofile;
|
| 1018 |
ofile.open (ss.str().c_str());
|
| 1019 |
ofile << res->getComment() << asctime (timeinfo)
|
| 1020 |
<< res->getComment()<< "\n"
|
| 1021 |
<< *res;
|
| 1022 |
ofile.close();
|
| 1023 |
delete res;
|
| 1024 |
}
|
