| 1 |
csander |
1.1 |
#include "TROOT.h"
|
| 2 |
|
|
#include "TSystem.h"
|
| 3 |
|
|
#include "TF1.h"
|
| 4 |
|
|
#include "TH1.h"
|
| 5 |
|
|
#include "TH2.h"
|
| 6 |
|
|
#include "TProfile.h"
|
| 7 |
|
|
#include "TCanvas.h"
|
| 8 |
|
|
#include "TRandom.h"
|
| 9 |
|
|
#include "TStyle.h"
|
| 10 |
|
|
#include "TObjArray.h"
|
| 11 |
|
|
#include "TGraph.h"
|
| 12 |
|
|
#include "TMath.h"
|
| 13 |
|
|
#include "TSpline.h"
|
| 14 |
|
|
|
| 15 |
|
|
#include <iostream>
|
| 16 |
|
|
#include <vector>
|
| 17 |
|
|
#include <string>
|
| 18 |
|
|
|
| 19 |
|
|
#include "external.h"
|
| 20 |
|
|
|
| 21 |
|
|
using namespace::std;
|
| 22 |
|
|
|
| 23 |
|
|
////////////////////////////////////////////////////////////
|
| 24 |
|
|
class TDataPoint{
|
| 25 |
|
|
|
| 26 |
|
|
private:
|
| 27 |
|
|
double weight;
|
| 28 |
|
|
double weight_orig;
|
| 29 |
|
|
double truth;
|
| 30 |
|
|
double meas;
|
| 31 |
|
|
double error;
|
| 32 |
|
|
|
| 33 |
|
|
public:
|
| 34 |
|
|
double GetTruth(){return truth;};
|
| 35 |
|
|
double GetMeas(){return meas;};
|
| 36 |
|
|
double GetError(){return error;};
|
| 37 |
|
|
double GetWeight(){return weight;};
|
| 38 |
|
|
double GetWeight_orig(){return weight_orig;};
|
| 39 |
|
|
void SetWeight(double w){weight=w;};
|
| 40 |
|
|
void SetWeight_orig(double w){weight_orig=w;};
|
| 41 |
|
|
|
| 42 |
|
|
//constructor
|
| 43 |
|
|
TDataPoint(double t, double m, double e, double w){
|
| 44 |
|
|
weight = w;
|
| 45 |
|
|
weight_orig = w;
|
| 46 |
|
|
truth = t;
|
| 47 |
|
|
meas = m;
|
| 48 |
|
|
error = e;
|
| 49 |
|
|
};
|
| 50 |
|
|
|
| 51 |
|
|
//destructor
|
| 52 |
|
|
~TDataPoint(){};
|
| 53 |
|
|
|
| 54 |
|
|
};
|
| 55 |
|
|
////////////////////////////////////////////////////////////
|
| 56 |
|
|
|
| 57 |
|
|
////////////////////////////////////////////////////////////
|
| 58 |
|
|
class TData{
|
| 59 |
|
|
|
| 60 |
|
|
private:
|
| 61 |
|
|
int Nevts;
|
| 62 |
|
|
double minX, maxX;
|
| 63 |
|
|
double minY, maxY;
|
| 64 |
|
|
double dY;
|
| 65 |
|
|
vector<TDataPoint> data;
|
| 66 |
|
|
vector<double> sp;
|
| 67 |
|
|
|
| 68 |
|
|
//----------------------------------------------------------
|
| 69 |
|
|
double chi2(vector<double>* PAR){
|
| 70 |
|
|
|
| 71 |
|
|
TGraph cgr(sp.size(), &sp.front(), &PAR->front());
|
| 72 |
|
|
TSpline3 *cspl = new TSpline3("cspl",&cgr);
|
| 73 |
|
|
double x2 = 0.;
|
| 74 |
|
|
for (vector<TDataPoint>::iterator i = data.begin(); i < data.end(); ++i){
|
| 75 |
|
|
double mess=i->GetMeas();
|
| 76 |
|
|
double truth=i->GetTruth();
|
| 77 |
|
|
/*
|
| 78 |
|
|
double c = calib(mess,cspl);
|
| 79 |
|
|
double dc = dcalib(mess,cspl);
|
| 80 |
|
|
double d = c * mess - truth;
|
| 81 |
|
|
double e = dc * i->GetError();
|
| 82 |
|
|
*/
|
| 83 |
|
|
/*
|
| 84 |
|
|
double c = calib(truth,cspl);
|
| 85 |
|
|
double d = c * mess - truth;
|
| 86 |
|
|
double e = c * i->GetError();
|
| 87 |
|
|
*/
|
| 88 |
|
|
double cmess = calib(mess,cspl);
|
| 89 |
|
|
double cold = 1;
|
| 90 |
|
|
double cinv = calibinv(truth,cmess,cold,cspl);
|
| 91 |
|
|
double d = mess - truth/cinv;
|
| 92 |
|
|
double e = i->GetError();
|
| 93 |
|
|
x2 += i->GetWeight()*pow(d/e, 2);
|
| 94 |
|
|
}
|
| 95 |
|
|
return x2;
|
| 96 |
|
|
}
|
| 97 |
|
|
//----------------------------------------------------------
|
| 98 |
|
|
|
| 99 |
|
|
//----------------------------------------------------------
|
| 100 |
|
|
double calib(double &x, TSpline3* calibSpline){
|
| 101 |
|
|
return calibSpline->Eval(x);
|
| 102 |
|
|
}
|
| 103 |
|
|
//----------------------------------------------------------
|
| 104 |
|
|
|
| 105 |
|
|
//----------------------------------------------------------
|
| 106 |
|
|
//might be usefull if calibration constant is defined as function of truth
|
| 107 |
|
|
double calibinv(double& x, double &cmess, double &cold, TSpline3* calibSpline){
|
| 108 |
|
|
double cinv = calibSpline->Eval(x/cmess);
|
| 109 |
|
|
//cout<<"cold: "<<cold<<" cnew: "<<cinv<<endl;
|
| 110 |
|
|
double delta = fabs((cinv-cold)/cold);
|
| 111 |
|
|
if (delta<1.e-4){
|
| 112 |
|
|
return cinv;
|
| 113 |
|
|
} else {
|
| 114 |
|
|
return calibSpline->Eval(x/calibinv(x,cmess,cinv,calibSpline));
|
| 115 |
|
|
}
|
| 116 |
|
|
}
|
| 117 |
|
|
//----------------------------------------------------------
|
| 118 |
|
|
|
| 119 |
|
|
//----------------------------------------------------------
|
| 120 |
|
|
double dcalib(double &mess, TSpline3* calibSpline){
|
| 121 |
|
|
double eps=1.;
|
| 122 |
|
|
double dc=((mess+eps)*calibSpline->Eval(mess+eps)-
|
| 123 |
|
|
(mess-eps)*calibSpline->Eval(mess-eps))/2./eps;
|
| 124 |
|
|
return dc;
|
| 125 |
|
|
}
|
| 126 |
|
|
//----------------------------------------------------------
|
| 127 |
|
|
|
| 128 |
|
|
//----------------------------------------------------------
|
| 129 |
|
|
//error parametrization, width of measured distribution
|
| 130 |
|
|
double error(double x){
|
| 131 |
|
|
double error = -1;
|
| 132 |
|
|
if ( x >= 0 ) error = 1.25*sqrt(x);
|
| 133 |
|
|
return error;
|
| 134 |
|
|
}
|
| 135 |
|
|
//----------------------------------------------------------
|
| 136 |
|
|
|
| 137 |
|
|
//----------------------------------------------------------
|
| 138 |
|
|
//reweight events to have flat truth distribution
|
| 139 |
|
|
void FlattenSpectra(){
|
| 140 |
|
|
|
| 141 |
|
|
TH1F* hist1 = new TH1F("truth non weighted","True distribution of measurement", 50, 0, 1000);
|
| 142 |
|
|
TH1F* hist2 = new TH1F("truth weighted","Weighted true distribution of measurement", 50, 0, 1000);
|
| 143 |
|
|
TH1F* weights = new TH1F("weights","Weights to flatten distribution", 50, 0, 1000);
|
| 144 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 145 |
|
|
hist1->Fill(i->GetTruth(),i->GetWeight());
|
| 146 |
|
|
}
|
| 147 |
|
|
double Integral1=hist1->Integral();
|
| 148 |
|
|
for (int i=0; i<hist1->GetNbinsX(); ++i){
|
| 149 |
|
|
weights->SetBinContent(i,1./hist1->GetBinContent(i));
|
| 150 |
|
|
}
|
| 151 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 152 |
|
|
double w=weights->GetBinContent(hist1->FindBin(i->GetTruth()));
|
| 153 |
|
|
hist2->Fill(i->GetTruth(),w);
|
| 154 |
|
|
}
|
| 155 |
|
|
double Integral2=hist2->Integral();
|
| 156 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 157 |
|
|
double w=weights->GetBinContent(hist1->FindBin(i->GetTruth()));
|
| 158 |
|
|
w*=i->GetWeight();
|
| 159 |
|
|
w*=Integral1/Integral2;
|
| 160 |
|
|
i->SetWeight(w);
|
| 161 |
|
|
i->SetWeight_orig(w);
|
| 162 |
|
|
}
|
| 163 |
|
|
}
|
| 164 |
|
|
//----------------------------------------------------------
|
| 165 |
|
|
|
| 166 |
|
|
//----------------------------------------------------------
|
| 167 |
|
|
//reweight spectra to correct for energy dependent resolution
|
| 168 |
|
|
void ResolutionReweight(bool logscale = false, bool outlier=false, int sigmamax=3){
|
| 169 |
|
|
|
| 170 |
|
|
cout<<"Events at start of ResolutionReweight: "<<data.size()<<endl;
|
| 171 |
|
|
|
| 172 |
|
|
int Nbins=49;
|
| 173 |
|
|
int minEntries=200;
|
| 174 |
|
|
vector<TDataPoint> dataCut;
|
| 175 |
|
|
vector<double> mean;
|
| 176 |
|
|
vector<double> sigma;
|
| 177 |
|
|
|
| 178 |
|
|
double d;
|
| 179 |
|
|
if (logscale){
|
| 180 |
|
|
d=(log(maxY)-log(minY))/Nbins;
|
| 181 |
|
|
} else {
|
| 182 |
|
|
d=(maxY-minY)/Nbins;
|
| 183 |
|
|
}
|
| 184 |
|
|
|
| 185 |
|
|
//book histos
|
| 186 |
|
|
TH1F* hist[Nbins];
|
| 187 |
|
|
TF1* f[Nbins];
|
| 188 |
|
|
for (int n=0; n<Nbins; ++n){
|
| 189 |
|
|
char hname[100];
|
| 190 |
|
|
sprintf(hname,"hr%i",n);
|
| 191 |
|
|
hist[n] = new TH1F(hname,"Truth for MeasBin", 1000, minX, maxX);
|
| 192 |
|
|
hist[n]->Sumw2();
|
| 193 |
|
|
}
|
| 194 |
|
|
|
| 195 |
|
|
//Fill histos
|
| 196 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 197 |
|
|
int n;
|
| 198 |
|
|
if (logscale){
|
| 199 |
|
|
n=(int)(Nbins*(log(i->GetMeas())-log(minY))/(log(maxY)-log(minY)));
|
| 200 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 201 |
|
|
} else {
|
| 202 |
|
|
n=(int)(Nbins*(i->GetMeas()-minY)/d);
|
| 203 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 204 |
|
|
};
|
| 205 |
|
|
hist[n]->Fill(i->GetTruth(),i->GetWeight());
|
| 206 |
|
|
}
|
| 207 |
|
|
|
| 208 |
|
|
//Fit histos
|
| 209 |
|
|
for (int n=0; n<Nbins; ++n){
|
| 210 |
|
|
cout<<n<<" of "<<Nbins<<endl;
|
| 211 |
|
|
if (hist[n]->GetEntries()<minEntries){
|
| 212 |
|
|
mean.push_back(0);
|
| 213 |
|
|
sigma.push_back(0);
|
| 214 |
|
|
continue;
|
| 215 |
|
|
}
|
| 216 |
|
|
char fname[100];
|
| 217 |
|
|
sprintf(fname,"fr%i",n);
|
| 218 |
|
|
f[n] = new TF1(fname,"gausn",minX,maxX);
|
| 219 |
|
|
//use a guess for starting values
|
| 220 |
|
|
f[n]->SetParameters(hist[n]->Integral("width"),hist[n]->GetMean(),hist[n]->GetRMS());
|
| 221 |
|
|
hist[n]->Fit(fname,"ILLRQN");
|
| 222 |
|
|
double Mean=f[n]->GetParameter(1);
|
| 223 |
|
|
double Sigma=f[n]->GetParameter(2);
|
| 224 |
|
|
mean.push_back(Mean);
|
| 225 |
|
|
sigma.push_back(Sigma);
|
| 226 |
|
|
}
|
| 227 |
|
|
|
| 228 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 229 |
|
|
int n;
|
| 230 |
|
|
if (logscale){
|
| 231 |
|
|
n=(int)(Nbins*(log(i->GetMeas())-log(minY))/(log(maxY)-log(minY)));
|
| 232 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 233 |
|
|
} else {
|
| 234 |
|
|
n=(int)(Nbins*(i->GetMeas()-minY)/d);
|
| 235 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 236 |
|
|
};
|
| 237 |
|
|
double delta = mean.at(n) - i->GetTruth();
|
| 238 |
|
|
//Use only events with truth value in 3 sigma of mean truth... ???
|
| 239 |
|
|
if (outlier && fabs(delta/sigma.at(n))>3.) continue;
|
| 240 |
|
|
double A, B;
|
| 241 |
|
|
double ERR = error(i->GetTruth()+2*delta);
|
| 242 |
|
|
if (ERR > 0){
|
| 243 |
|
|
A = 1./sqrt(2*TMath::Pi())/i->GetError()*exp(-pow(delta,2)/2./pow(i->GetError(),2));
|
| 244 |
|
|
B = 1./sqrt(2*TMath::Pi())/ERR*exp(-pow(delta,2)/2./pow(ERR,2));
|
| 245 |
|
|
double newweight = 2*B/(A+B);
|
| 246 |
|
|
i->SetWeight(i->GetWeight_orig()*newweight);
|
| 247 |
|
|
}
|
| 248 |
|
|
dataCut.push_back(*i);
|
| 249 |
|
|
}
|
| 250 |
|
|
data.clear();
|
| 251 |
|
|
data.assign(dataCut.begin(),dataCut.end());
|
| 252 |
|
|
cout<<"Events at end of ResolutionReweight: "<<data.size()<<endl;
|
| 253 |
|
|
|
| 254 |
|
|
minY=99999.;
|
| 255 |
|
|
maxY=0;
|
| 256 |
|
|
minX=99999.;
|
| 257 |
|
|
maxX=0;
|
| 258 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 259 |
|
|
if (i->GetMeas() < minY) minY=i->GetMeas();
|
| 260 |
|
|
if (i->GetMeas() > maxY) maxY=i->GetMeas();
|
| 261 |
|
|
if (i->GetTruth() < minX) minX=i->GetTruth();
|
| 262 |
|
|
if (i->GetTruth() > maxX) maxX=i->GetTruth();
|
| 263 |
|
|
}
|
| 264 |
|
|
|
| 265 |
|
|
}
|
| 266 |
|
|
//----------------------------------------------------------
|
| 267 |
|
|
|
| 268 |
|
|
//----------------------------------------------------------
|
| 269 |
|
|
//symmetrize truth distribution for one measurement
|
| 270 |
|
|
void CorrectCutoff(bool logscale = false){
|
| 271 |
|
|
|
| 272 |
|
|
cout<<"Events at start of CorrectCutoff: "<<data.size()<<endl;
|
| 273 |
|
|
int Nbins=49;
|
| 274 |
|
|
int minEntries=200;
|
| 275 |
|
|
vector<TDataPoint> dataCut;
|
| 276 |
|
|
vector<double> mean;
|
| 277 |
|
|
vector<double> dmean;
|
| 278 |
|
|
|
| 279 |
|
|
TCanvas *c1 = new TCanvas("c1","Correct for cut off",0,0,1000,1000);
|
| 280 |
|
|
c1->Divide(7,7);
|
| 281 |
|
|
double d;
|
| 282 |
|
|
if (logscale){
|
| 283 |
|
|
d=(log(maxY)-log(minY))/Nbins;
|
| 284 |
|
|
} else {
|
| 285 |
|
|
d=(maxY-minY)/Nbins;
|
| 286 |
|
|
}
|
| 287 |
|
|
|
| 288 |
|
|
//book histos
|
| 289 |
|
|
TH1F* hist[Nbins];
|
| 290 |
|
|
TF1* f[Nbins];
|
| 291 |
|
|
for (int n=0; n<Nbins; ++n){
|
| 292 |
|
|
char hname[100];
|
| 293 |
|
|
sprintf(hname,"hc%i",n);
|
| 294 |
|
|
hist[n] = new TH1F(hname,"Truth for MeasBin", 1000, minX, maxX);
|
| 295 |
|
|
hist[n]->Sumw2();
|
| 296 |
|
|
}
|
| 297 |
|
|
|
| 298 |
|
|
//Fill histos
|
| 299 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 300 |
|
|
int n;
|
| 301 |
|
|
if (logscale){
|
| 302 |
|
|
n=(int)(Nbins*(log(i->GetMeas())-log(minY))/(log(maxY)-log(minY)));
|
| 303 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 304 |
|
|
} else {
|
| 305 |
|
|
n=(int)(Nbins*(i->GetMeas()-minY)/d);
|
| 306 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 307 |
|
|
};
|
| 308 |
|
|
hist[n]->Fill(i->GetTruth(),i->GetWeight());
|
| 309 |
|
|
}
|
| 310 |
|
|
|
| 311 |
|
|
//Fit histos
|
| 312 |
|
|
for (int n=0; n<Nbins; ++n){
|
| 313 |
|
|
cout<<n<<" of "<<Nbins<<endl;
|
| 314 |
|
|
if (hist[n]->GetEntries()<minEntries){
|
| 315 |
|
|
mean.push_back(0);
|
| 316 |
|
|
dmean.push_back(0);
|
| 317 |
|
|
continue;
|
| 318 |
|
|
}
|
| 319 |
|
|
char fname[100];
|
| 320 |
|
|
sprintf(fname,"fc%i",n);
|
| 321 |
|
|
f[n] = new TF1(fname,"gausn",minX,maxX);
|
| 322 |
|
|
//use a guess for starting values
|
| 323 |
|
|
f[n]->SetParameters(hist[n]->Integral("width"),hist[n]->GetMean(),hist[n]->GetRMS());
|
| 324 |
|
|
c1->cd(n+1);
|
| 325 |
|
|
hist[n]->SetLineWidth(0);
|
| 326 |
|
|
hist[n]->Draw();
|
| 327 |
|
|
f[n]->SetLineWidth(0);
|
| 328 |
|
|
f[n]->SetLineColor(2);
|
| 329 |
|
|
hist[n]->Fit(fname,"ILLRQ");
|
| 330 |
|
|
double Mean=f[n]->GetParameter(1);
|
| 331 |
|
|
mean.push_back(Mean);
|
| 332 |
|
|
double dMean=0;
|
| 333 |
|
|
if (Mean<minX || Mean>maxX){
|
| 334 |
|
|
dMean=0;
|
| 335 |
|
|
} else {
|
| 336 |
|
|
double dMeanMin=fabs(Mean-minX);
|
| 337 |
|
|
double dMeanMax=fabs(Mean-maxX);
|
| 338 |
|
|
dMeanMin<dMeanMax?dMean=dMeanMin:dMean=dMeanMax;
|
| 339 |
|
|
}
|
| 340 |
|
|
dmean.push_back(dMean);
|
| 341 |
|
|
}
|
| 342 |
|
|
c1->SaveAs("cut_LVMini.eps");
|
| 343 |
|
|
|
| 344 |
|
|
//symmetrize distributions
|
| 345 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 346 |
|
|
int n;
|
| 347 |
|
|
if (logscale){
|
| 348 |
|
|
n=(int)(Nbins*(log(i->GetMeas())-log(minY))/(log(maxY)-log(minY)));
|
| 349 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 350 |
|
|
} else {
|
| 351 |
|
|
n=(int)(Nbins*(i->GetMeas()-minY)/d);
|
| 352 |
|
|
if (n>=Nbins) n=Nbins-1;
|
| 353 |
|
|
};
|
| 354 |
|
|
double delta = fabs(mean.at(n) - i->GetTruth());
|
| 355 |
|
|
if (delta>dmean.at(n)) continue;
|
| 356 |
|
|
dataCut.push_back(*i);
|
| 357 |
|
|
}
|
| 358 |
|
|
|
| 359 |
|
|
data.clear();
|
| 360 |
|
|
data.assign(dataCut.begin(),dataCut.end());
|
| 361 |
|
|
cout<<"Events at end of CorrectCutoff: "<<data.size()<<endl;
|
| 362 |
|
|
|
| 363 |
|
|
minY=99999.;
|
| 364 |
|
|
maxY=0;
|
| 365 |
|
|
minX=99999.;
|
| 366 |
|
|
maxX=0;
|
| 367 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 368 |
|
|
if (i->GetMeas() < minY) minY=i->GetMeas();
|
| 369 |
|
|
if (i->GetMeas() > maxY) maxY=i->GetMeas();
|
| 370 |
|
|
if (i->GetTruth() < minX) minX=i->GetTruth();
|
| 371 |
|
|
if (i->GetTruth() > maxX) maxX=i->GetTruth();
|
| 372 |
|
|
}
|
| 373 |
|
|
|
| 374 |
|
|
}
|
| 375 |
|
|
//----------------------------------------------------------
|
| 376 |
|
|
|
| 377 |
|
|
public:
|
| 378 |
|
|
|
| 379 |
|
|
vector<TDataPoint>* GetDataRef(){return &data;};
|
| 380 |
|
|
|
| 381 |
|
|
//----------------------------------------------------------
|
| 382 |
|
|
void DrawTruth(){
|
| 383 |
|
|
TH1F* hist = new TH1F("truth","True distribution of measurement", 100, 0, maxX);
|
| 384 |
|
|
hist->Sumw2();
|
| 385 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 386 |
|
|
hist->Fill(i->GetTruth(),i->GetWeight());
|
| 387 |
|
|
}
|
| 388 |
|
|
TCanvas *c1 = new TCanvas("c1","Truth",0,0,600,600);
|
| 389 |
|
|
c1->Divide(1,1);
|
| 390 |
|
|
c1->cd(1);
|
| 391 |
|
|
c1->cd(1)->SetLogy();
|
| 392 |
|
|
hist->SetXTitle("Truth");
|
| 393 |
|
|
hist->Draw();
|
| 394 |
|
|
c1->SaveAs("truth_LVMini.eps");
|
| 395 |
|
|
};
|
| 396 |
|
|
//----------------------------------------------------------
|
| 397 |
|
|
|
| 398 |
|
|
//----------------------------------------------------------
|
| 399 |
|
|
void DrawMeas(){
|
| 400 |
|
|
TH1F* hist = new TH1F("meas","Measured distribution of measurement", 100, 0, maxY);
|
| 401 |
|
|
hist->Sumw2();
|
| 402 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 403 |
|
|
hist->Fill(i->GetMeas(),i->GetWeight());
|
| 404 |
|
|
}
|
| 405 |
|
|
TCanvas *c1 = new TCanvas("c2","Measurement",0,0,600,600);
|
| 406 |
|
|
c1->Divide(1,1);
|
| 407 |
|
|
c1->cd(1);
|
| 408 |
|
|
c1->cd(1)->SetLogy();
|
| 409 |
|
|
hist->SetXTitle("Measurement");
|
| 410 |
|
|
hist->Draw();
|
| 411 |
|
|
c1->SaveAs("meas_LVMini.eps");
|
| 412 |
|
|
};
|
| 413 |
|
|
//----------------------------------------------------------
|
| 414 |
|
|
|
| 415 |
|
|
//----------------------------------------------------------
|
| 416 |
|
|
void DrawCalibConst(vector<double> p){
|
| 417 |
|
|
TGraph cgr(sp.size(), &sp.front(), &p.front());
|
| 418 |
|
|
TSpline3 *cspl = new TSpline3("cspl",&cgr);
|
| 419 |
|
|
vector<double> vx, vy;
|
| 420 |
|
|
int Nbins = 100;
|
| 421 |
|
|
for (int i=0; i<=Nbins; ++i){
|
| 422 |
|
|
double x=minY+i*(maxY-minY)/Nbins;
|
| 423 |
|
|
vx.push_back(double(x));
|
| 424 |
|
|
double c=calib(x,cspl);
|
| 425 |
|
|
vy.push_back(c);
|
| 426 |
|
|
}
|
| 427 |
|
|
TGraph* CalibConst = new TGraph(vx.size(), &vx.front(), &vy.front());
|
| 428 |
|
|
TCanvas *c1 = new TCanvas("c3","Calibration Constant vs. Measurement",0,0,600,600);
|
| 429 |
|
|
c1->Divide(1,1);
|
| 430 |
|
|
c1->cd(1);
|
| 431 |
|
|
CalibConst->SetLineColor(2);
|
| 432 |
|
|
CalibConst->SetLineWidth(4);
|
| 433 |
|
|
CalibConst->SetTitle("Calibration Constant");
|
| 434 |
|
|
CalibConst->GetXaxis()->SetTitle("Truth");
|
| 435 |
|
|
CalibConst->GetYaxis()->SetTitle("C");
|
| 436 |
|
|
CalibConst->SetMinimum(0.9);
|
| 437 |
|
|
CalibConst->SetMaximum(1.3);
|
| 438 |
|
|
CalibConst->Draw("ACP");
|
| 439 |
|
|
//CalibConst->Draw("L");
|
| 440 |
|
|
c1->SaveAs("calibconst_LVMini.eps");
|
| 441 |
|
|
};
|
| 442 |
|
|
//----------------------------------------------------------
|
| 443 |
|
|
|
| 444 |
|
|
//----------------------------------------------------------
|
| 445 |
|
|
void DrawRatio(vector<double> p){
|
| 446 |
|
|
TGraph cgr(sp.size(), &sp.front(), &p.front());
|
| 447 |
|
|
TSpline3 *cspl = new TSpline3("cspl",&cgr);
|
| 448 |
|
|
TH2F* hist1 = new TH2F("r1","Measurement/Truth vs. Truth", 100, 0, 1000, 100, 0.0, 2.0);
|
| 449 |
|
|
TH2F* hist2 = new TH2F("r2","Measurement/Truth vs. Measurement", 100, 0, 1000, 100, 0.0, 2.0);
|
| 450 |
|
|
TH2F* hist3 = new TH2F("r3","Measurement/Truth vs. CalibMeasurement", 100, 0, 1000, 100, 0.0, 2.0);
|
| 451 |
|
|
TH2F* hist4 = new TH2F("r4","CalibMeasurement/Truth vs. Truth", 100, 0, 1000, 100, 0.0, 2.0);
|
| 452 |
|
|
TH2F* hist5 = new TH2F("r5","CalibMeasurement/Truth vs. Measurement", 100, 0, 1000, 100, 0.0, 2.0);
|
| 453 |
|
|
TH2F* hist6 = new TH2F("r6","CalibMeasurement/Truth vs. CalibMeasurement", 100, 0, 1000, 100, 0.0, 2.0);
|
| 454 |
|
|
hist1->Sumw2();
|
| 455 |
|
|
hist2->Sumw2();
|
| 456 |
|
|
hist3->Sumw2();
|
| 457 |
|
|
hist4->Sumw2();
|
| 458 |
|
|
hist5->Sumw2();
|
| 459 |
|
|
hist6->Sumw2();
|
| 460 |
|
|
hist1->SetTitleOffset(1.2,"Y");
|
| 461 |
|
|
hist2->SetTitleOffset(1.2,"Y");
|
| 462 |
|
|
hist3->SetTitleOffset(1.2,"Y");
|
| 463 |
|
|
hist4->SetTitleOffset(1.2,"Y");
|
| 464 |
|
|
hist5->SetTitleOffset(1.2,"Y");
|
| 465 |
|
|
hist6->SetTitleOffset(1.2,"Y");
|
| 466 |
|
|
hist1->SetTitleOffset(1.2,"Y");
|
| 467 |
|
|
hist1->SetYTitle("k=Meas/Truth");
|
| 468 |
|
|
hist2->SetYTitle("k=Meas/Truth");
|
| 469 |
|
|
hist3->SetYTitle("k=Meas/Truth");
|
| 470 |
|
|
hist4->SetYTitle("k=calibMeas/Truth");
|
| 471 |
|
|
hist5->SetYTitle("k=calibMeas/Truth");
|
| 472 |
|
|
hist6->SetYTitle("k=calibMeas/Truth");
|
| 473 |
|
|
hist1->SetXTitle("Truth");
|
| 474 |
|
|
hist4->SetXTitle("Truth");
|
| 475 |
|
|
hist2->SetXTitle("Measurement");
|
| 476 |
|
|
hist5->SetXTitle("Measurement");
|
| 477 |
|
|
hist3->SetXTitle("calib. Measurement");
|
| 478 |
|
|
hist6->SetXTitle("calib. Measurement");
|
| 479 |
|
|
for (vector<TDataPoint>::iterator i=data.begin(); i<data.end(); ++i){
|
| 480 |
|
|
double mess=i->GetMeas();
|
| 481 |
|
|
double truth=i->GetTruth();
|
| 482 |
|
|
double c=calib(mess,cspl);
|
| 483 |
|
|
hist1->Fill(truth, mess/truth, i->GetWeight());
|
| 484 |
|
|
hist2->Fill(mess, mess/truth, i->GetWeight());
|
| 485 |
|
|
hist3->Fill(c*mess, mess/truth, i->GetWeight());
|
| 486 |
|
|
hist4->Fill(truth, (c*mess)/truth, i->GetWeight());
|
| 487 |
|
|
hist5->Fill(mess, (c*mess)/truth, i->GetWeight());
|
| 488 |
|
|
hist6->Fill(c*mess, (c*mess)/truth, i->GetWeight());
|
| 489 |
|
|
}
|
| 490 |
|
|
gROOT->SetStyle("Plain");
|
| 491 |
|
|
gStyle->SetPalette(51,0);
|
| 492 |
|
|
TCanvas *c1 = new TCanvas("c4","Ration",0,0,900,900);
|
| 493 |
|
|
TF1 *f = new TF1("f","gaus",minX,maxX);
|
| 494 |
|
|
TObjArray aSlices;
|
| 495 |
|
|
c1->Divide(3,3);
|
| 496 |
|
|
c1->cd(1);
|
| 497 |
|
|
gPad->SetRightMargin(0.15);
|
| 498 |
|
|
gPad->SetLeftMargin(0.15);
|
| 499 |
|
|
hist1->Draw("COLZ");
|
| 500 |
|
|
hist1->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 501 |
|
|
TH1* hist1_1 = (TH1*)(aSlices[1]);
|
| 502 |
|
|
TH1* hist1_2 = (TH1*)(aSlices[2]);
|
| 503 |
|
|
for (int i=0; i<hist1_1->GetNbinsX(); ++i){
|
| 504 |
|
|
hist1_1->SetBinError(i,hist1_2->GetBinContent(i));
|
| 505 |
|
|
}
|
| 506 |
|
|
hist1_1->SetLineColor(2);
|
| 507 |
|
|
hist1_1->Draw("same");
|
| 508 |
|
|
hist1->SetLineColor(5);
|
| 509 |
|
|
hist1->ProfileX()->Draw("same");
|
| 510 |
|
|
|
| 511 |
|
|
c1->cd(2);
|
| 512 |
|
|
gPad->SetRightMargin(0.12);
|
| 513 |
|
|
hist2->Draw("COLZ");
|
| 514 |
|
|
hist2->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 515 |
|
|
TH1* hist2_1 = (TH1*)(aSlices[1]);
|
| 516 |
|
|
TH1* hist2_2 = (TH1*)(aSlices[2]);
|
| 517 |
|
|
for (int i=0; i<hist2_1->GetNbinsX(); ++i){
|
| 518 |
|
|
hist2_1->SetBinError(i,hist2_2->GetBinContent(i));
|
| 519 |
|
|
}
|
| 520 |
|
|
hist2_1->SetLineColor(2);
|
| 521 |
|
|
hist2_1->Draw("same");
|
| 522 |
|
|
hist2->SetLineColor(5);
|
| 523 |
|
|
hist2->ProfileX()->Draw("same");
|
| 524 |
|
|
|
| 525 |
|
|
c1->cd(3);
|
| 526 |
|
|
gPad->SetRightMargin(0.12);
|
| 527 |
|
|
hist3->Draw("COLZ");
|
| 528 |
|
|
hist3->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 529 |
|
|
TH1* hist3_1 = (TH1*)(aSlices[1]);
|
| 530 |
|
|
TH1* hist3_2 = (TH1*)(aSlices[2]);
|
| 531 |
|
|
for (int i=0; i<hist3_1->GetNbinsX(); ++i){
|
| 532 |
|
|
hist3_1->SetBinError(i,hist3_2->GetBinContent(i));
|
| 533 |
|
|
}
|
| 534 |
|
|
hist3_1->SetLineColor(2);
|
| 535 |
|
|
hist3_1->Draw("same");
|
| 536 |
|
|
hist3->SetLineColor(5);
|
| 537 |
|
|
hist3->ProfileX()->Draw("same");
|
| 538 |
|
|
|
| 539 |
|
|
c1->cd(4);
|
| 540 |
|
|
gPad->SetRightMargin(0.12);
|
| 541 |
|
|
hist4->Draw("COLZ");
|
| 542 |
|
|
hist4->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 543 |
|
|
TH1* hist4_1 = (TH1*)(aSlices[1]);
|
| 544 |
|
|
TH1* hist4_2 = (TH1*)(aSlices[2]);
|
| 545 |
|
|
for (int i=0; i<hist4_1->GetNbinsX(); ++i){
|
| 546 |
|
|
hist4_1->SetBinError(i,hist4_2->GetBinContent(i));
|
| 547 |
|
|
}
|
| 548 |
|
|
hist4_1->SetLineColor(2);
|
| 549 |
|
|
hist4_1->Draw("same");
|
| 550 |
|
|
hist4->SetLineColor(5);
|
| 551 |
|
|
hist4->ProfileX()->Draw("same");
|
| 552 |
|
|
|
| 553 |
|
|
c1->cd(5);
|
| 554 |
|
|
gPad->SetRightMargin(0.12);
|
| 555 |
|
|
hist5->Draw("COLZ");
|
| 556 |
|
|
hist5->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 557 |
|
|
TH1* hist5_1 = (TH1*)(aSlices[1]);
|
| 558 |
|
|
TH1* hist5_2 = (TH1*)(aSlices[2]);
|
| 559 |
|
|
for (int i=0; i<hist5_1->GetNbinsX(); ++i){
|
| 560 |
|
|
hist5_1->SetBinError(i,hist5_2->GetBinContent(i));
|
| 561 |
|
|
}
|
| 562 |
|
|
hist5_1->SetLineColor(2);
|
| 563 |
|
|
hist5_1->Draw("same");
|
| 564 |
|
|
hist5->SetLineColor(5);
|
| 565 |
|
|
hist5->ProfileX()->Draw("same");
|
| 566 |
|
|
|
| 567 |
|
|
c1->cd(6);
|
| 568 |
|
|
gPad->SetRightMargin(0.12);
|
| 569 |
|
|
hist6->Draw("COLZ");
|
| 570 |
|
|
hist6->FitSlicesY(f, 0, -1, 0, "NQ", &aSlices);
|
| 571 |
|
|
TH1* hist6_1 = (TH1*)(aSlices[1]);
|
| 572 |
|
|
TH1* hist6_2 = (TH1*)(aSlices[2]);
|
| 573 |
|
|
for (int i=0; i<hist6_1->GetNbinsX(); ++i){
|
| 574 |
|
|
hist6_1->SetBinError(i,hist6_2->GetBinContent(i));
|
| 575 |
|
|
}
|
| 576 |
|
|
hist6_1->SetLineColor(2);
|
| 577 |
|
|
hist6_1->Draw("same");
|
| 578 |
|
|
hist6->SetLineColor(5);
|
| 579 |
|
|
hist6->ProfileX()->Draw("same");
|
| 580 |
|
|
|
| 581 |
|
|
c1->cd(7);
|
| 582 |
|
|
hist4_1->SetTitleOffset(1.2,"Y");
|
| 583 |
|
|
hist4_1->SetXTitle("Truth");
|
| 584 |
|
|
hist4_1->SetYTitle("k");
|
| 585 |
|
|
hist4_1->SetLineWidth(0);
|
| 586 |
|
|
hist4_1->SetLineColor(2);
|
| 587 |
|
|
hist4_1->SetMinimum(0.95);
|
| 588 |
|
|
hist4_1->SetMaximum(1.05);
|
| 589 |
|
|
hist4_1->Draw();
|
| 590 |
|
|
|
| 591 |
|
|
c1->cd(8);
|
| 592 |
|
|
hist5_1->SetTitleOffset(1.2,"Y");
|
| 593 |
|
|
hist5_1->SetXTitle("Measurement");
|
| 594 |
|
|
hist5_1->SetYTitle("k");
|
| 595 |
|
|
hist5_1->SetLineWidth(0);
|
| 596 |
|
|
hist5_1->SetLineColor(2);
|
| 597 |
|
|
hist5_1->SetMinimum(0.95);
|
| 598 |
|
|
hist5_1->SetMaximum(1.05);
|
| 599 |
|
|
hist5_1->Draw();
|
| 600 |
|
|
|
| 601 |
|
|
c1->cd(9);
|
| 602 |
|
|
hist6_1->SetTitleOffset(1.2,"Y");
|
| 603 |
|
|
hist6_1->SetXTitle("calib. Measurement");
|
| 604 |
|
|
hist6_1->SetYTitle("k");
|
| 605 |
|
|
hist6_1->SetLineWidth(0);
|
| 606 |
|
|
hist6_1->SetLineColor(2);
|
| 607 |
|
|
hist6_1->SetMinimum(0.95);
|
| 608 |
|
|
hist6_1->SetMaximum(1.05);
|
| 609 |
|
|
hist6_1->Draw();
|
| 610 |
|
|
|
| 611 |
|
|
c1->SaveAs("ratio_LVMini.eps");
|
| 612 |
|
|
};
|
| 613 |
|
|
//----------------------------------------------------------
|
| 614 |
|
|
|
| 615 |
|
|
//----------------------------------------------------------
|
| 616 |
|
|
//constructor
|
| 617 |
|
|
TData(int N=100, double min=300., double max=700., int npar=10,
|
| 618 |
|
|
bool flat=false,
|
| 619 |
|
|
bool cutoff=false, bool cutofflog=false,
|
| 620 |
|
|
bool reweight=false, bool reweightlog=false,
|
| 621 |
|
|
bool reco=false){
|
| 622 |
|
|
Nevts = N;
|
| 623 |
|
|
minX=99999;
|
| 624 |
|
|
maxX=0;
|
| 625 |
|
|
minY=99990;
|
| 626 |
|
|
maxY=0;
|
| 627 |
|
|
int Nnegative=0;
|
| 628 |
|
|
TF1 *recoeff = new TF1("f","pow(x/30.,2)/(1.+pow(x/30.,2))");
|
| 629 |
|
|
for (int i=0;i<Nevts;i++) {
|
| 630 |
|
|
double x = gRandom->Rndm(i);
|
| 631 |
|
|
//double truth = min+(max-min)*x;
|
| 632 |
|
|
double truth = gRandom->Exp(150);
|
| 633 |
|
|
double width = error(truth);
|
| 634 |
|
|
double meas = gRandom->Gaus(0.9*truth,width);
|
| 635 |
|
|
double pr = gRandom->Rndm(i);
|
| 636 |
|
|
//double pr = -9999.;
|
| 637 |
|
|
if (meas<0 && truth>min){
|
| 638 |
|
|
++Nnegative;
|
| 639 |
|
|
cout<<"WARNING: negative measurement (truth="<<truth<<",meas="<<meas<<") no. "<<Nnegative<<endl;
|
| 640 |
|
|
}
|
| 641 |
|
|
if (meas>0 && truth>min && pr<(*recoeff)(truth)){
|
| 642 |
|
|
if (meas<minY) minY=meas;
|
| 643 |
|
|
if (meas>maxY) maxY=meas;
|
| 644 |
|
|
if (truth<minX) minX=truth;
|
| 645 |
|
|
if (truth>maxX) maxX=truth;
|
| 646 |
|
|
double error = width;
|
| 647 |
|
|
double weight=1.;
|
| 648 |
|
|
//reco eff correction
|
| 649 |
|
|
if (reco) weight = 1./(*recoeff)(truth);
|
| 650 |
|
|
TDataPoint d(truth, meas, error, weight);
|
| 651 |
|
|
data.push_back(d);
|
| 652 |
|
|
}
|
| 653 |
|
|
}
|
| 654 |
|
|
sp.clear();
|
| 655 |
|
|
|
| 656 |
|
|
if (flat) FlattenSpectra();
|
| 657 |
|
|
if (reweight){
|
| 658 |
|
|
bool outlier=true;
|
| 659 |
|
|
ResolutionReweight(reweightlog, outlier, 5);
|
| 660 |
|
|
ResolutionReweight(reweightlog, outlier, 4);
|
| 661 |
|
|
ResolutionReweight(reweightlog, outlier, 3);
|
| 662 |
|
|
}
|
| 663 |
|
|
if (cutoff) CorrectCutoff(cutofflog);
|
| 664 |
|
|
|
| 665 |
|
|
dY=(maxY-minY)/(npar-3);
|
| 666 |
|
|
for (int i=0; i<npar; ++i){
|
| 667 |
|
|
sp.push_back(minY+(i-1)*dY);
|
| 668 |
|
|
}
|
| 669 |
|
|
|
| 670 |
|
|
};
|
| 671 |
|
|
//----------------------------------------------------------
|
| 672 |
|
|
|
| 673 |
|
|
//----------------------------------------------------------
|
| 674 |
|
|
void evalF(int &NPAR, vector<double>* GRAD, double &FSUM, vector<double>* PAR){
|
| 675 |
|
|
|
| 676 |
|
|
FSUM = chi2(PAR);
|
| 677 |
|
|
// gradients
|
| 678 |
|
|
double eps=1.e-5;
|
| 679 |
|
|
for (int i=0; i<NPAR; ++i){
|
| 680 |
|
|
vector<double> PAR1 = *PAR;
|
| 681 |
|
|
vector<double> PAR2 = *PAR;
|
| 682 |
|
|
PAR1.at(i) += eps;
|
| 683 |
|
|
PAR2.at(i) -= eps;
|
| 684 |
|
|
GRAD->at(i) = (chi2(&PAR1)-chi2(&PAR2))/2./eps;
|
| 685 |
|
|
PAR1.at(i) += eps;
|
| 686 |
|
|
PAR2.at(i) -= eps;
|
| 687 |
|
|
GRAD->at(i+NPAR) = (chi2(&PAR1)-2*FSUM+chi2(&PAR2))/4./eps/eps;
|
| 688 |
|
|
}
|
| 689 |
|
|
}
|
| 690 |
|
|
//----------------------------------------------------------
|
| 691 |
|
|
|
| 692 |
|
|
//destructor
|
| 693 |
|
|
~TData(){};
|
| 694 |
|
|
|
| 695 |
|
|
};
|
| 696 |
|
|
////////////////////////////////////////////////////////////
|
| 697 |
|
|
|
| 698 |
|
|
////////////////////////////////////////////////////////////
|
| 699 |
|
|
int main(){
|
| 700 |
|
|
|
| 701 |
|
|
int NPAR=11;
|
| 702 |
|
|
|
| 703 |
|
|
TROOT simple("simple","Some calibration toyMC tests");
|
| 704 |
|
|
bool flat = true;
|
| 705 |
|
|
bool cutoff = true;
|
| 706 |
|
|
bool cutofflog = true;
|
| 707 |
|
|
bool reweight = true;
|
| 708 |
|
|
bool reweightlog = true;
|
| 709 |
|
|
bool reco = true;
|
| 710 |
|
|
TData* toy= new TData(1000000,20,99999,NPAR,flat,cutoff,cutofflog,reweight,reweightlog,reco);
|
| 711 |
|
|
toy->DrawTruth();
|
| 712 |
|
|
toy->DrawMeas();
|
| 713 |
|
|
|
| 714 |
|
|
////// LVMINI //////
|
| 715 |
|
|
int NITER = 500;
|
| 716 |
|
|
int MVEC;
|
| 717 |
|
|
NPAR<29?MVEC=NPAR:MVEC=29;
|
| 718 |
|
|
int NAUX = 10000;
|
| 719 |
|
|
|
| 720 |
|
|
vector<double> AUX(NAUX, 0.);
|
| 721 |
|
|
vector<double> parameter;
|
| 722 |
|
|
double FSUM;
|
| 723 |
|
|
double FOPT, FEDM, DUMMY;
|
| 724 |
|
|
|
| 725 |
|
|
for (int i=0; i<NPAR; ++i){
|
| 726 |
|
|
parameter.push_back(1./0.9);
|
| 727 |
|
|
}
|
| 728 |
|
|
float eps =float(1.E-3);
|
| 729 |
|
|
float wlf1=1.E-4;
|
| 730 |
|
|
float wlf2=0.9;
|
| 731 |
|
|
//default values
|
| 732 |
|
|
//double wlf1=-1.;
|
| 733 |
|
|
//double wlf2=-1.;
|
| 734 |
|
|
|
| 735 |
|
|
lvmeps_(eps,wlf1,wlf2);
|
| 736 |
|
|
|
| 737 |
|
|
int NPARNEG = -fabs(NPAR);
|
| 738 |
|
|
if(lvmdim_(NPAR, MVEC) > NAUX)
|
| 739 |
|
|
cout<<"Aux field too small: "<<NAUX<<"<"<<lvmdim_(NPAR, MVEC)<<" needed entries"<<endl;
|
| 740 |
|
|
lvmini_(NPARNEG, MVEC, NITER, &AUX.front());
|
| 741 |
|
|
int IFLAG;
|
| 742 |
|
|
int IRET = 0;
|
| 743 |
|
|
int ITER = 0;
|
| 744 |
|
|
do {
|
| 745 |
|
|
++ITER;
|
| 746 |
|
|
toy->evalF(NPAR, &AUX, FSUM, ¶meter);
|
| 747 |
|
|
lvmfun_(¶meter.front(), FSUM, IRET, &AUX.front());
|
| 748 |
|
|
lvmprt_(2, &AUX.front(), 2); //print out
|
| 749 |
|
|
/*
|
| 750 |
|
|
if (ITER%10==0){
|
| 751 |
|
|
toy->DrawRatio(parameter);
|
| 752 |
|
|
toy->DrawCalibConst(parameter);
|
| 753 |
|
|
}
|
| 754 |
|
|
*/
|
| 755 |
|
|
} while (IRET<0);
|
| 756 |
|
|
int error_index=2;
|
| 757 |
|
|
error_index = lvmind_(error_index);
|
| 758 |
|
|
|
| 759 |
|
|
toy->DrawRatio(parameter);
|
| 760 |
|
|
toy->DrawCalibConst(parameter);
|
| 761 |
|
|
|
| 762 |
|
|
return 0;
|
| 763 |
|
|
|
| 764 |
|
|
}
|
| 765 |
|
|
////////////////////////////////////////////////////////////
|
| 766 |
|
|
|
