Logo ROOT   6.30.04
Reference Guide
 All Namespaces Files Pages
HFitInterface.cxx
Go to the documentation of this file.
1 // @(#)root/hist:$Id$
2 // Author: L. Moneta Thu Aug 31 10:40:20 2006
3 
4 /**********************************************************************
5  * *
6  * Copyright (c) 2006 LCG ROOT Math Team, CERN/PH-SFT *
7  * *
8  * *
9  **********************************************************************/
10 
11 // Implementation file for class TH1Interface
12 
13 #include "HFitInterface.h"
14 
15 #include "Fit/BinData.h"
16 #include "Fit/SparseData.h"
17 #include "Fit/FitResult.h"
18 #include "Math/IParamFunction.h"
19 
20 #include <vector>
21 
22 #include <cassert>
23 #include <cmath>
24 
25 #include "TH1.h"
26 #include "THnBase.h"
27 #include "TF1.h"
28 #include "TGraph2D.h"
29 #include "TGraph.h"
30 #include "TGraphErrors.h"
31 // #include "TGraphErrors.h"
32 // #include "TGraphBentErrors.h"
33 // #include "TGraphAsymmErrors.h"
34 #include "TMultiGraph.h"
35 #include "TList.h"
36 #include "TError.h"
37 
38 
39 //#define DEBUG
40 #ifdef DEBUG
41 #include "TClass.h"
42 #include <iostream>
43 #endif
44 
45 
46 namespace ROOT {
47 
48 namespace Fit {
49 
50 // add a namespace to distinguish from the Graph functions
51 namespace HFitInterface {
52 
53 
54 bool IsPointOutOfRange(const TF1 * func, const double * x) {
55  // function to check if a point is outside range
56  if (func ==0) return false;
57  return !func->IsInside(x);
58 }
59 
60 bool AdjustError(const DataOptions & option, double & error, double value = 1) {
61  // adjust the given error according to the option
62  // return false when point must be skipped.
63  // When point error = 0, the point is kept if the option UseEmpty is set or if
64  // fErrors1 is set and the point value is not zero.
65  // The value should be used only for points representing counts (histograms), not for the graph.
66  // In the graph points with zero errors are by default skipped indepentently of the value.
67  // If one wants to keep the points, the option fUseEmpty must be set
68 
69  if (error <= 0) {
70  if (option.fUseEmpty || (option.fErrors1 && std::abs(value) > 0 ) )
71  error = 1.; // set error to 1
72  else
73  return false; // skip bins with zero errors or empty
74  } else if (option.fErrors1)
75  error = 1; // set all error to 1 for non-empty bins
76  return true;
77 }
78 
79 void ExamineRange(const TAxis * axis, std::pair<double,double> range,int &hxfirst,int &hxlast) {
80  // examine the range given with the pair on the given histogram axis
81  // correct in case the bin values hxfirst hxlast
82  double xlow = range.first;
83  double xhigh = range.second;
84 #ifdef DEBUG
85  std::cout << "xlow " << xlow << " xhigh = " << xhigh << std::endl;
86 #endif
87  // ignore ranges specified outside histogram range
88  int ilow = axis->FindFixBin(xlow);
89  int ihigh = axis->FindFixBin(xhigh);
90  if (ilow > hxlast || ihigh < hxfirst) {
91  Warning("ROOT::Fit::FillData","fit range is outside histogram range, no fit data for %s",axis->GetName());
92  }
93  // consider only range defined with-in histogram not oustide. Always exclude underflow/overflow
94  hxfirst = std::min( std::max( ilow, hxfirst), hxlast+1) ;
95  hxlast = std::max( std::min( ihigh, hxlast), hxfirst-1) ;
96  // exclude bins where range coverage is less than half bin width
97  if (hxfirst < hxlast) {
98  if ( axis->GetBinCenter(hxfirst) < xlow) hxfirst++;
99  if ( axis->GetBinCenter(hxlast) > xhigh) hxlast--;
100  }
101 }
102 
103 
104 } // end namespace HFitInterface
105 
106 
107 void FillData(BinData & dv, const TH1 * hfit, TF1 * func)
108 {
109  // Function to fill the binned Fit data structure from a TH1
110  // The dimension of the data is the same of the histogram dimension
111  // The function pointer is need in case of integral is used and to reject points
112  // rejected in the function
113 
114  // the TF1 pointer cannot be constant since EvalPar and InitArgs are not const methods
115 
116  // get fit option
117  const DataOptions & fitOpt = dv.Opt();
118 
119 
120  // store instead of bin center the bin edges
121  bool useBinEdges = fitOpt.fIntegral || fitOpt.fBinVolume;
122 
123  assert(hfit != 0);
124 
125  //std::cout << "creating Fit Data from histogram " << hfit->GetName() << std::endl;
126 
127  int hxfirst = hfit->GetXaxis()->GetFirst();
128  int hxlast = hfit->GetXaxis()->GetLast();
129 
130  int hyfirst = hfit->GetYaxis()->GetFirst();
131  int hylast = hfit->GetYaxis()->GetLast();
132 
133  int hzfirst = hfit->GetZaxis()->GetFirst();
134  int hzlast = hfit->GetZaxis()->GetLast();
135 
136  // function by default has same range (use that one if requested otherwise use data one)
137 
138 
139  // get the range (add the function range ??)
140  // to check if inclusion/exclusion at end/point
141  const DataRange & range = dv.Range();
142  if (range.Size(0) != 0) {
143  HFitInterface::ExamineRange( hfit->GetXaxis(), range(0), hxfirst, hxlast);
144  if (range.Size(0) > 1 ) {
145  Warning("ROOT::Fit::FillData","support only one range interval for X coordinate");
146  }
147  }
148 
149  if (hfit->GetDimension() > 1 && range.Size(1) != 0) {
150  HFitInterface::ExamineRange( hfit->GetYaxis(), range(1), hyfirst, hylast);
151  if (range.Size(1) > 1 )
152  Warning("ROOT::Fit::FillData","support only one range interval for Y coordinate");
153  }
154 
155  if (hfit->GetDimension() > 2 && range.Size(2) != 0) {
156  HFitInterface::ExamineRange( hfit->GetZaxis(), range(2), hzfirst, hzlast);
157  if (range.Size(2) > 1 )
158  Warning("ROOT::Fit::FillData","support only one range interval for Z coordinate");
159  }
160 
161 
162  int n = (hxlast-hxfirst+1)*(hylast-hyfirst+1)*(hzlast-hzfirst+1);
163 
164 #ifdef DEBUG
165  std::cout << "THFitInterface: ifirst = " << hxfirst << " ilast = " << hxlast
166  << " total bins " << n
167  << std::endl;
168 #endif
169 
170  // reserve n for more efficient usage
171  //dv.Data().reserve(n);
172 
173  int hdim = hfit->GetDimension();
174  int ndim = hdim;
175  // case of function dimension less than histogram
176  if (func !=0 && func->GetNdim() == hdim-1) ndim = hdim-1;
177 
178  assert( ndim > 0 );
179  //typedef BinPoint::CoordData CoordData;
180  //CoordData x = CoordData( hfit->GetDimension() );
181  dv.Initialize(n,ndim);
182 
183  double x[3];
184  double s[3];
185 
186  int binx = 0;
187  int biny = 0;
188  int binz = 0;
189 
190  const TAxis *xaxis = hfit->GetXaxis();
191  const TAxis *yaxis = hfit->GetYaxis();
192  const TAxis *zaxis = hfit->GetZaxis();
193 
194  for ( binx = hxfirst; binx <= hxlast; ++binx) {
195  if (useBinEdges) {
196  x[0] = xaxis->GetBinLowEdge(binx);
197  s[0] = xaxis->GetBinUpEdge(binx);
198  }
199  else
200  x[0] = xaxis->GetBinCenter(binx);
201 
202 
203  for ( biny = hyfirst; biny <= hylast; ++biny) {
204  if (useBinEdges) {
205  x[1] = yaxis->GetBinLowEdge(biny);
206  s[1] = yaxis->GetBinUpEdge(biny);
207  }
208  else
209  x[1] = yaxis->GetBinCenter(biny);
210 
211  for ( binz = hzfirst; binz <= hzlast; ++binz) {
212  if (useBinEdges) {
213  x[2] = zaxis->GetBinLowEdge(binz);
214  s[2] = zaxis->GetBinUpEdge(binz);
215  }
216  else
217  x[2] = zaxis->GetBinCenter(binz);
218 
219  // need to evaluate function to know about rejected points
220  // hugly but no other solutions
221  if (func != 0) {
222  TF1::RejectPoint(false);
223  (*func)( &x[0] ); // evaluate using stored function parameters
224  if (TF1::RejectedPoint() ) continue;
225  }
226 
227 
228  double value = hfit->GetBinContent(binx, biny, binz);
229  double error = hfit->GetBinError(binx, biny, binz);
230  if (!HFitInterface::AdjustError(fitOpt,error,value) ) continue;
231 
232  if (ndim == hdim -1) {
233  // case of fitting a function with dimension -1
234  // point error is bin width y / sqrt(N) where N is nuber of entries in the bin
235  // normalization of error will be wrong - but they will be rescaled in the fit
236  if (hdim == 2) dv.Add( x, x[1], yaxis->GetBinWidth(biny) / error );
237  if (hdim == 3) dv.Add( x, x[2], zaxis->GetBinWidth(binz) / error );
238  } else {
239  dv.Add( x, value, error );
240  if (useBinEdges) {
241  dv.AddBinUpEdge( s );
242  }
243  }
244 
245 
246 #ifdef DEBUG
247  std::cout << "bin " << binx << " add point " << x[0] << " " << hfit->GetBinContent(binx) << std::endl;
248 #endif
249 
250  } // end loop on z bins
251  } // end loop on y bins
252  } // end loop on x axis
253 
254 
255 #ifdef DEBUG
256  std::cout << "THFitInterface::FillData: Hist FitData size is " << dv.Size() << std::endl;
257 #endif
258 
259 }
260 
261 ////////////////////////////////////////////////////////////////////////////////
262 /// Compute rough values of parameters for an exponential
263 
264 void InitExpo(const ROOT::Fit::BinData & data, TF1 * f1)
265 {
266  unsigned int n = data.Size();
267  if (n == 0) return;
268 
269  // find xmin and xmax of the data
270  double valxmin;
271  double xmin = *(data.GetPoint(0,valxmin));
272  double xmax = xmin;
273  double valxmax = valxmin;
274 
275  for (unsigned int i = 1; i < n; ++ i) {
276  double val;
277  double x = *(data.GetPoint(i,val) );
278  if (x < xmin) {
279  xmin = x;
280  valxmin = val;
281  }
282  else if (x > xmax) {
283  xmax = x;
284  valxmax = val;
285  }
286  }
287 
288  // avoid negative values of valxmin/valxmax
289  if (valxmin <= 0 && valxmax > 0 ) valxmin = valxmax;
290  else if (valxmax <=0 && valxmin > 0) valxmax = valxmin;
291  else if (valxmin <=0 && valxmax <= 0) { valxmin = 1; valxmax = 1; }
292 
293  double slope = std::log( valxmax/valxmin) / (xmax - xmin);
294  double constant = std::log(valxmin) - slope * xmin;
295  f1->SetParameters(constant, slope);
296 }
297 
298 
299 ////////////////////////////////////////////////////////////////////////////////
300 /// Compute Initial values of parameters for a gaussian
301 /// derived from function H1InitGaus defined in TH1.cxx
302 
303 void InitGaus(const ROOT::Fit::BinData & data, TF1 * f1)
304 {
305 
306  static const double sqrtpi = 2.506628;
307 
308  // - Compute mean value and RMS of the data
309  unsigned int n = data.Size();
310  if (n == 0) return;
311  double sumx = 0;
312  double sumx2 = 0;
313  double allcha = 0;
314  double valmax = 0;
315  double rangex = data.Coords(n-1)[0] - data.Coords(0)[0];
316  // to avoid binwidth = 0 set arbitrarly to 1
317  double binwidth = 1;
318  if ( rangex > 0) binwidth = rangex;
319  double x0 = 0;
320  for (unsigned int i = 0; i < n; ++ i) {
321  double val;
322  double x = *(data.GetPoint(i,val) );
323  sumx += val*x;
324  sumx2 += val*x*x;
325  allcha += val;
326  if (val > valmax) valmax = val;
327  if (i > 0) {
328  double dx = x - x0;
329  if (dx < binwidth) binwidth = dx;
330  }
331  x0 = x;
332  }
333 
334  if (allcha <= 0) return;
335  double mean = sumx/allcha;
336  double rms = sumx2/allcha - mean*mean;
337 
338 
339  if (rms > 0)
340  rms = std::sqrt(rms);
341  else
342  rms = binwidth*n/4;
343 
344 
345  //if the distribution is really gaussian, the best approximation
346  //is binwidx*allcha/(sqrtpi*rms)
347  //However, in case of non-gaussian tails, this underestimates
348  //the normalisation constant. In this case the maximum value
349  //is a better approximation.
350  //We take the average of both quantities
351 
352 // printf("valmax %f other %f bw %f allcha %f rms %f \n",valmax, binwidth*allcha/(sqrtpi*rms),
353 // binwidth, allcha,rms );
354 
355  double constant = 0.5*(valmax+ binwidth*allcha/(sqrtpi*rms));
356 
357 
358  //In case the mean value is outside the histo limits and
359  //the RMS is bigger than the range, we take
360  // mean = center of bins
361  // rms = half range
362 // Double_t xmin = curHist->GetXaxis()->GetXmin();
363 // Double_t xmax = curHist->GetXaxis()->GetXmax();
364 // if ((mean < xmin || mean > xmax) && rms > (xmax-xmin)) {
365 // mean = 0.5*(xmax+xmin);
366 // rms = 0.5*(xmax-xmin);
367 // }
368 
369  f1->SetParameter(0,constant);
370  f1->SetParameter(1,mean);
371  f1->SetParameter(2,rms);
372  f1->SetParLimits(2,0,10*rms);
373 
374 
375 #ifdef DEBUG
376  std::cout << "Gaussian initial par values" << constant << " " << mean << " " << rms << std::endl;
377 #endif
378 
379 }
380 
381 ////////////////////////////////////////////////////////////////////////////////
382 /// Compute Initial values of parameters for a gaussian
383 /// derived from function H1InitGaus defined in TH1.cxx
384 
385 void Init2DGaus(const ROOT::Fit::BinData & data, TF1 * f1)
386 {
387 
388  static const double sqrtpi = 2.506628;
389 
390  // - Compute mean value and RMS of the data
391  unsigned int n = data.Size();
392  if (n == 0) return;
393  double sumx = 0, sumy = 0;
394  double sumx2 = 0, sumy2 = 0;
395  double allcha = 0;
396  double valmax = 0;
397  double rangex = data.Coords(n-1)[0] - data.Coords(0)[0];
398  double rangey = data.Coords(n-1)[1] - data.Coords(0)[1];
399  // to avoid binwidthx = 0 set arbitrarly to 1
400  double binwidthx = 1, binwidthy = 1;
401  if ( rangex > 0) binwidthx = rangex;
402  if ( rangey > 0) binwidthy = rangey;
403  double x0 = 0, y0 = 0;
404  for (unsigned int i = 0; i < n; ++i) {
405  double val;
406  const double *coords = data.GetPoint(i,val);
407  double x = coords[0], y = coords[1];
408  sumx += val*x;
409  sumy += val*y;
410  sumx2 += val*x*x;
411  sumy2 += val*y*y;
412  allcha += val;
413  if (val > valmax) valmax = val;
414  if (i > 0) {
415  double dx = x - x0;
416  if (dx < binwidthx) binwidthx = dx;
417  double dy = y - y0;
418  if (dy < binwidthy) binwidthy = dy;
419  }
420  x0 = x;
421  y0 = y;
422  }
423 
424  if (allcha <= 0) return;
425  double meanx = sumx/allcha, meany = sumy/allcha;
426  double rmsx = sumx2/allcha - meanx*meanx;
427  double rmsy = sumy2/allcha - meany*meany;
428 
429 
430  if (rmsx > 0)
431  rmsx = std::sqrt(rmsx);
432  else
433  rmsx = binwidthx*n/4;
434 
435  if (rmsy > 0)
436  rmsy = std::sqrt(rmsy);
437  else
438  rmsy = binwidthy*n/4;
439 
440 
441  //if the distribution is really gaussian, the best approximation
442  //is binwidx*allcha/(sqrtpi*rmsx)
443  //However, in case of non-gaussian tails, this underestimates
444  //the normalisation constant. In this case the maximum value
445  //is a better approximation.
446  //We take the average of both quantities
447 
448  double constant = 0.5 * (valmax+ binwidthx*allcha/(sqrtpi*rmsx))*
449  (valmax+ binwidthy*allcha/(sqrtpi*rmsy));
450 
451  f1->SetParameter(0,constant);
452  f1->SetParameter(1,meanx);
453  f1->SetParameter(2,rmsx);
454  f1->SetParLimits(2,0,10*rmsx);
455  f1->SetParameter(3,meany);
456  f1->SetParameter(4,rmsy);
457  f1->SetParLimits(4,0,10*rmsy);
458 
459 #ifdef DEBUG
460  std::cout << "2D Gaussian initial par values"
461  << constant << " "
462  << meanx << " "
463  << rmsx
464  << meany << " "
465  << rmsy
466  << std::endl;
467 #endif
468 
469 }
470 
471 // filling fit data from TGraph objects
472 
473 BinData::ErrorType GetDataType(const TGraph * gr, DataOptions & fitOpt) {
474  // get type of data for TGraph objects
475  double *ex = gr->GetEX();
476  double *ey = gr->GetEY();
477  double * eyl = gr->GetEYlow();
478  double * eyh = gr->GetEYhigh();
479 
480 
481  // default case for graphs (when they have errors)
482  BinData::ErrorType type = BinData::kValueError;
483  // if all errors are zero set option of using errors to 1
484  if (fitOpt.fErrors1 || ( ey == 0 && ( eyl == 0 || eyh == 0 ) ) ) {
485  type = BinData::kNoError;
486  }
487  // need to treat case when all errors are zero
488  // note that by default fitOpt.fCoordError is true
489  else if ( ex != 0 && fitOpt.fCoordErrors) {
490  // check that all errors are not zero
491  int i = 0;
492  while (i < gr->GetN() && type != BinData::kCoordError) {
493  if (ex[i] > 0) type = BinData::kCoordError;
494  ++i;
495  }
496  }
497  // case of asymmetric errors (by default fAsymErrors is true)
498  else if ( ( eyl != 0 && eyh != 0) && fitOpt.fAsymErrors) {
499  // check also if that all errors are non zero's
500  int i = 0;
501  bool zeroErrorX = true;
502  bool zeroErrorY = true;
503  while (i < gr->GetN() && (zeroErrorX || zeroErrorY)) {
504  double e2X = ( gr->GetErrorXlow(i) + gr->GetErrorXhigh(i) );
505  double e2Y = eyl[i] + eyh[i];
506  zeroErrorX &= (e2X <= 0);
507  zeroErrorY &= (e2Y <= 0);
508  ++i;
509  }
510  if (zeroErrorX && zeroErrorY)
511  type = BinData::kNoError;
512  else if (!zeroErrorX && zeroErrorY)
513  type = BinData::kCoordError;
514  else if (zeroErrorX && !zeroErrorY) {
515  type = BinData::kAsymError;
516  fitOpt.fCoordErrors = false;
517  }
518  else {
519  type = BinData::kAsymError;
520  }
521  }
522 
523  // need to look also a case when all errors in y are zero
524  if ( ey != 0 && type != BinData::kCoordError ) {
525  int i = 0;
526  bool zeroError = true;
527  while (i < gr->GetN() && zeroError) {
528  if (ey[i] > 0) zeroError = false;;
529  ++i;
530  }
531  if (zeroError) type = BinData::kNoError;
532  }
533 
534 
535 #ifdef DEBUG
536  std::cout << "type is " << type << " graph type is " << gr->IsA()->GetName() << std::endl;
537 #endif
538 
539  return type;
540 }
541 
542 BinData::ErrorType GetDataType(const TGraph2D * gr, const DataOptions & fitOpt) {
543  // get type of data for TGraph2D object
544  double *ex = gr->GetEX();
545  double *ey = gr->GetEY();
546  double *ez = gr->GetEZ();
547 
548  // default case for graphs (when they have errors)
549  BinData::ErrorType type = BinData::kValueError;
550  // if all errors are zero set option of using errors to 1
551  if (fitOpt.fErrors1 || ez == 0 ) {
552  type = BinData::kNoError;
553  }
554  else if ( ex != 0 && ey!=0 && fitOpt.fCoordErrors) {
555  // check that all errors are not zero
556  int i = 0;
557  while (i < gr->GetN() && type != BinData::kCoordError) {
558  if (ex[i] > 0 || ey[i] > 0) type = BinData::kCoordError;
559  ++i;
560  }
561  }
562 
563 
564 #ifdef DEBUG
565  std::cout << "type is " << type << " graph2D type is " << gr->IsA()->GetName() << std::endl;
566 #endif
567 
568  return type;
569 }
570 
571 
572 
573 void DoFillData ( BinData & dv, const TGraph * gr, BinData::ErrorType type, TF1 * func ) {
574  // internal method to do the actual filling of the data
575  // given a graph and a multigraph
576 
577  // get fit option
578  DataOptions & fitOpt = dv.Opt();
579 
580  int nPoints = gr->GetN();
581  double *gx = gr->GetX();
582  double *gy = gr->GetY();
583 
584  const DataRange & range = dv.Range();
585  bool useRange = ( range.Size(0) > 0);
586  double xmin = 0;
587  double xmax = 0;
588  range.GetRange(xmin,xmax);
589 
590  dv.Initialize(nPoints,1, type);
591 
592 #ifdef DEBUG
593  std::cout << "DoFillData: graph npoints = " << nPoints << " type " << type << std::endl;
594  if (func) {
595  double a1,a2; func->GetRange(a1,a2); std::cout << "func range " << a1 << " " << a2 << std::endl;
596  }
597 #endif
598 
599  double x[1];
600  for ( int i = 0; i < nPoints; ++i) {
601 
602  x[0] = gx[i];
603 
604 
605  if (useRange && ( x[0] < xmin || x[0] > xmax) ) continue;
606 
607  // need to evaluate function to know about rejected points
608  // hugly but no other solutions
609  if (func) {
610  TF1::RejectPoint(false);
611  (*func)( x ); // evaluate using stored function parameters
612  if (TF1::RejectedPoint() ) continue;
613  }
614 
615 
616  if (fitOpt.fErrors1)
617  dv.Add( gx[i], gy[i] );
618 
619  // for the errors use the getters by index to avoid cases when the arrays are zero
620  // (like in a case of a graph)
621  else if (type == BinData::kValueError) {
622  double errorY = gr->GetErrorY(i);
623  // should consider error = 0 as 1 ? Decide to skip points with zero errors
624  // in case want to keep points with error = 0 as errrors=1 need to set the option UseEmpty
625  if (!HFitInterface::AdjustError(fitOpt,errorY) ) continue;
626  dv.Add( gx[i], gy[i], errorY );
627 
628 #ifdef DEBUG
629  std::cout << "Point " << i << " " << gx[i] << " " << gy[i] << " " << errorY << std::endl;
630 #endif
631 
632 
633  }
634  else { // case use error in x or asym errors
635  double errorX = 0;
636  if (fitOpt.fCoordErrors)
637  // shoulkd take combined average (sqrt(0.5(e1^2+e2^2)) or math average ?
638  // gr->GetErrorX(i) returns combined average
639  // use math average for same behaviour as before
640  errorX = std::max( 0.5 * ( gr->GetErrorXlow(i) + gr->GetErrorXhigh(i) ) , 0. ) ;
641 
642 
643  // adjust error in y according to option
644  double errorY = std::max(gr->GetErrorY(i), 0.);
645  // we do not check the return value since we check later if error in X and Y is zero for skipping the point
646  HFitInterface::AdjustError(fitOpt, errorY);
647 
648  // skip points with total error = 0
649  if ( errorX <=0 && errorY <= 0 ) continue;
650 
651 
652  if (type == BinData::kAsymError) {
653  // asymmetric errors
654  dv.Add( gx[i], gy[i], errorX, gr->GetErrorYlow(i), gr->GetErrorYhigh(i) );
655  }
656  else {
657  // case symmetric Y errors
658  dv.Add( gx[i], gy[i], errorX, errorY );
659  }
660  }
661 
662  }
663 
664 #ifdef DEBUG
665  std::cout << "TGraphFitInterface::FillData Graph FitData size is " << dv.Size() << std::endl;
666 #endif
667 
668 }
669 
670 void FillData(SparseData & dv, const TH1 * h1, TF1 * /*func*/)
671 {
672  const int dim = h1->GetDimension();
673  std::vector<double> min(dim);
674  std::vector<double> max(dim);
675 
676  int ncells = h1->GetNcells();
677  for ( int i = 0; i < ncells; ++i ) {
678 // printf("i: %d; OF: %d; UF: %d; C: %f\n"
679 // , i
680 // , h1->IsBinOverflow(i) , h1->IsBinUnderflow(i)
681 // , h1->GetBinContent(i));
682  if ( !( h1->IsBinOverflow(i) || h1->IsBinUnderflow(i) )
683  && h1->GetBinContent(i))
684  {
685  int x,y,z;
686  h1->GetBinXYZ(i, x, y, z);
687 
688 // std::cout << "FILLDATA: h1(" << i << ")"
689 // << "[" << h1->GetXaxis()->GetBinLowEdge(x) << "-" << h1->GetXaxis()->GetBinUpEdge(x) << "]";
690 // if ( dim >= 2 )
691 // std::cout << "[" << h1->GetYaxis()->GetBinLowEdge(y) << "-" << h1->GetYaxis()->GetBinUpEdge(y) << "]";
692 // if ( dim >= 3 )
693 // std::cout << "[" << h1->GetZaxis()->GetBinLowEdge(z) << "-" << h1->GetZaxis()->GetBinUpEdge(z) << "]";
694 
695 // std::cout << h1->GetBinContent(i) << std::endl;
696 
697  min[0] = h1->GetXaxis()->GetBinLowEdge(x);
698  max[0] = h1->GetXaxis()->GetBinUpEdge(x);
699  if ( dim >= 2 )
700  {
701  min[1] = h1->GetYaxis()->GetBinLowEdge(y);
702  max[1] = h1->GetYaxis()->GetBinUpEdge(y);
703  }
704  if ( dim >= 3 ) {
705  min[2] = h1->GetZaxis()->GetBinLowEdge(z);
706  max[2] = h1->GetZaxis()->GetBinUpEdge(z);
707  }
708 
709  dv.Add(min, max, h1->GetBinContent(i), h1->GetBinError(i));
710  }
711  }
712 }
713 
714 void FillData(SparseData & dv, const THnBase * h1, TF1 * /*func*/)
715 {
716  const int dim = h1->GetNdimensions();
717  std::vector<double> min(dim);
718  std::vector<double> max(dim);
719  std::vector<Int_t> coord(dim);
720 
721  ULong64_t nEntries = h1->GetNbins();
722  for ( ULong64_t i = 0; i < nEntries; i++ )
723  {
724  double value = h1->GetBinContent( i, &coord[0] );
725  if ( !value ) continue;
726 
727 // std::cout << "FILLDATA(SparseData): h1(" << i << ")";
728 
729  // Exclude underflows and oveflows! (defect behaviour with the TH1*)
730  bool insertBox = true;
731  for ( int j = 0; j < dim && insertBox; ++j )
732  {
733  TAxis* axis = h1->GetAxis(j);
734  if ( ( axis->GetBinLowEdge(coord[j]) < axis->GetXmin() ) ||
735  ( axis->GetBinUpEdge(coord[j]) > axis->GetXmax() ) ) {
736  insertBox = false;
737  }
738  min[j] = h1->GetAxis(j)->GetBinLowEdge(coord[j]);
739  max[j] = h1->GetAxis(j)->GetBinUpEdge(coord[j]);
740  }
741  if ( !insertBox ) {
742 // std::cout << "NOT INSERTED!"<< std::endl;
743  continue;
744  }
745 
746 // for ( int j = 0; j < dim; ++j )
747 // {
748 // std::cout << "[" << h1->GetAxis(j)->GetBinLowEdge(coord[j])
749 // << "-" << h1->GetAxis(j)->GetBinUpEdge(coord[j]) << "]";
750 // }
751 // std::cout << h1->GetBinContent(i) << std::endl;
752 
753  dv.Add(min, max, value, h1->GetBinError(i));
754  }
755 }
756 
757 void FillData(BinData & dv, const THnBase * s1, TF1 * func)
758 {
759  // Fill the Range of the THnBase
760  unsigned int const ndim = s1->GetNdimensions();
761  std::vector<double> xmin(ndim);
762  std::vector<double> xmax(ndim);
763  for ( unsigned int i = 0; i < ndim; ++i ) {
764  TAxis* axis = s1->GetAxis(i);
765  xmin[i] = axis->GetXmin();
766  xmax[i] = axis->GetXmax();
767  }
768 
769  // Put default options, needed for the likelihood fitting of sparse
770  // data.
771  ROOT::Fit::DataOptions& dopt = dv.Opt();
772  dopt.fUseEmpty = true;
773  // when using sparse data need to set option to use normalized bin volume, because sparse bins are merged together
774  //if (!dopt.fIntegral) dopt.fBinVolume = true;
775  dopt.fBinVolume = true;
776  dopt.fNormBinVolume = true;
777 
778  // Get the sparse data
779  ROOT::Fit::SparseData d(ndim, &xmin[0], &xmax[0]);
780  ROOT::Fit::FillData(d, s1, func);
781 
782 // std::cout << "FillData(BinData & dv, const THnBase * s1, TF1 * func) (1)" << std::endl;
783 
784  // Create the bin data from the sparse data
785  d.GetBinDataIntegral(dv);
786 
787 }
788 
789 void FillData ( BinData & dv, const TGraph * gr, TF1 * func ) {
790  // fill the data vector from a TGraph. Pass also the TF1 function which is
791  // needed in case to exclude points rejected by the function
792  assert(gr != 0);
793 
794  // get fit option
795  DataOptions & fitOpt = dv.Opt();
796 
797  BinData::ErrorType type = GetDataType(gr,fitOpt);
798  // adjust option according to type
799  fitOpt.fErrors1 = (type == BinData::kNoError);
800  // set this if we want to have error=1 for points with zero errors (by default they are skipped)
801  // fitOpt.fUseEmpty = true;
802 
803  // use coordinate or asym errors in case option is set and type is consistent
804  fitOpt.fCoordErrors &= (type == BinData::kCoordError) || (type == BinData::kAsymError) ;
805  fitOpt.fAsymErrors &= (type == BinData::kAsymError);
806 
807 
808  // if data are filled already check if there are consistent - otherwise do nothing
809  if (dv.Size() > 0 && dv.NDim() == 1 ) {
810  // check if size is correct otherwise flag an errors
811  if ( dv.GetErrorType() != type ) {
812  Error("FillData","Inconsistent TGraph with previous data set- skip all graph data");
813  return;
814  }
815  }
816 
817  DoFillData(dv, gr, type, func);
818 
819 }
820 
821 void FillData ( BinData & dv, const TMultiGraph * mg, TF1 * func ) {
822  // fill the data vector from a TMultiGraph. Pass also the TF1 function which is
823  // needed in case to exclude points rejected by the function
824  assert(mg != 0);
825 
826  TList * grList = mg->GetListOfGraphs();
827  assert(grList != 0);
828 
829 #ifdef DEBUG
830 // grList->Print();
831  TIter itr(grList, kIterBackward);
832  TObject *obj;
833  std::cout << "multi-graph list of graps: " << std::endl;
834  while ((obj = itr())) {
835  std::cout << obj->IsA()->GetName() << std::endl;
836  }
837 
838 #endif
839 
840  // get fit option
841  DataOptions & fitOpt = dv.Opt();
842 
843  // loop on the graphs to get the data type (use maximum)
844  TIter next(grList);
845 
846  BinData::ErrorType type = BinData::kNoError;
847  TGraph *gr = 0;
848  while ((gr = (TGraph*) next())) {
849  BinData::ErrorType t = GetDataType(gr,fitOpt);
850  if (t > type ) type = t;
851  }
852  // adjust option according to type
853  fitOpt.fErrors1 = (type == BinData::kNoError);
854  fitOpt.fCoordErrors = (type == BinData::kCoordError);
855  fitOpt.fAsymErrors = (type == BinData::kAsymError);
856 
857 
858 #ifdef DEBUG
859  std::cout << "Fitting MultiGraph of type " << type << std::endl;
860 #endif
861 
862  // fill the data now
863  next = grList;
864  while ((gr = (TGraph*) next())) {
865  DoFillData( dv, gr, type, func);
866  }
867 
868 #ifdef DEBUG
869  std::cout << "TGraphFitInterface::FillData MultiGraph FitData size is " << dv.Size() << std::endl;
870 #endif
871 
872 }
873 
874 void FillData ( BinData & dv, const TGraph2D * gr, TF1 * func ) {
875  // fill the data vector from a TGraph2D. Pass also the TF1 function which is
876  // needed in case to exclude points rejected by the function
877  // in case of a pure TGraph
878  assert(gr != 0);
879 
880  // get fit option
881  DataOptions & fitOpt = dv.Opt();
882  BinData::ErrorType type = GetDataType(gr,fitOpt);
883  // adjust option according to type
884  fitOpt.fErrors1 = (type == BinData::kNoError);
885  fitOpt.fCoordErrors = (type == BinData::kCoordError);
886  fitOpt.fAsymErrors = false; // a TGraph2D with asymmetric errors does not exist
887 
888  int nPoints = gr->GetN();
889  double *gx = gr->GetX();
890  double *gy = gr->GetY();
891  double *gz = gr->GetZ();
892 
893  // if all errors are zero set option of using errors to 1
894  if ( gr->GetEZ() == 0) fitOpt.fErrors1 = true;
895 
896  double x[2];
897  double ex[2];
898 
899  // look at data range
900  const DataRange & range = dv.Range();
901  bool useRangeX = ( range.Size(0) > 0);
902  bool useRangeY = ( range.Size(1) > 0);
903  double xmin = 0;
904  double xmax = 0;
905  double ymin = 0;
906  double ymax = 0;
907  range.GetRange(xmin,xmax,ymin,ymax);
908 
909  dv.Initialize(nPoints,2, type);
910 
911  for ( int i = 0; i < nPoints; ++i) {
912 
913  x[0] = gx[i];
914  x[1] = gy[i];
915 
916  //if (fitOpt.fUseRange && HFitInterface::IsPointOutOfRange(func, x) ) continue;
917  if (useRangeX && ( x[0] < xmin || x[0] > xmax) ) continue;
918  if (useRangeY && ( x[1] < ymin || x[1] > ymax) ) continue;
919 
920  // need to evaluate function to know about rejected points
921  // hugly but no other solutions
922  if (func) {
923  TF1::RejectPoint(false);
924  (*func)( x ); // evaluate using stored function parameters
925  if (TF1::RejectedPoint() ) continue;
926  }
927 
928  if (type == BinData::kNoError) {
929  dv.Add( x, gz[i] );
930  continue;
931  }
932 
933  double errorZ = gr->GetErrorZ(i);
934  if (!HFitInterface::AdjustError(fitOpt,errorZ) ) continue;
935 
936  if (type == BinData::kValueError) {
937  dv.Add( x, gz[i], errorZ );
938  }
939  else if (type == BinData::kCoordError) { // case use error in coordinates (x and y)
940  ex[0] = std::max(gr->GetErrorX(i), 0.);
941  ex[1] = std::max(gr->GetErrorY(i), 0.);
942  dv.Add( x, gz[i], ex, errorZ );
943  }
944  else
945  assert(0); // should not go here
946 
947 #ifdef DEBUG
948  std::cout << "Point " << i << " " << gx[i] << " " << gy[i] << " " << errorZ << std::endl;
949 #endif
950 
951  }
952 
953 #ifdef DEBUG
954  std::cout << "THFitInterface::FillData Graph2D FitData size is " << dv.Size() << std::endl;
955 #endif
956 
957 }
958 
959 
960 // confidence intervals
961 bool GetConfidenceIntervals(const TH1 * h1, const ROOT::Fit::FitResult & result, TGraphErrors * gr, double cl ) {
962  if (h1->GetDimension() != 1) {
963  Error("GetConfidenceIntervals","Invalid object used for storing confidence intervals");
964  return false;
965  }
966  // fill fit data sets with points to estimate cl.
967  BinData d;
968  FillData(d,h1,0);
969  gr->Set(d.NPoints() );
970  double * ci = gr->GetEY(); // make CL values error of the graph
971  result.GetConfidenceIntervals(d,ci,cl);
972  // put function value as abscissa of the graph
973  for (unsigned int ipoint = 0; ipoint < d.NPoints(); ++ipoint) {
974  const double * x = d.Coords(ipoint);
975  const ROOT::Math::IParamMultiFunction * func = result.FittedFunction();
976  gr->SetPoint(ipoint, x[0], (*func)(x) );
977  }
978  return true;
979 }
980 
981 
982 } // end namespace Fit
983 
984 } // end namespace ROOT
985