Logo ROOT   6.18/05
Reference Guide
rs_numberCountingCombination.C
Go to the documentation of this file.
1/// \file
2/// \ingroup tutorial_roostats
3/// \notebook -js
4/// 'Number Counting Example' RooStats tutorial macro #100
5///
6/// This tutorial shows an example of a combination of
7/// two searches using number counting with background uncertainty.
8///
9/// The macro uses a RooStats "factory" to construct a PDF
10/// that represents the two number counting analyses with background
11/// uncertainties. The uncertainties are taken into account by
12/// considering a sideband measurement of a size that corresponds to the
13/// background uncertainty. The problem has been studied in these references:
14/// - http://arxiv.org/abs/physics/0511028
15/// - http://arxiv.org/abs/physics/0702156
16/// - http://cdsweb.cern.ch/record/1099969?ln=en
17///
18/// After using the factory to make the model, we use a RooStats
19/// ProfileLikelihoodCalculator for a Hypothesis test and a confidence interval.
20/// The calculator takes into account systematics by eliminating nuisance parameters
21/// with the profile likelihood. This is equivalent to the method of MINOS.
22///
23///
24/// \macro_image
25/// \macro_output
26/// \macro_code
27///
28/// \author Kyle Cranmer
29
35#include "RooRealVar.h"
36
37#include <cassert>
38
39// use this order for safety on library loading
40using namespace RooFit;
41using namespace RooStats;
42
43// declare three variations on the same tutorial
44void rs_numberCountingCombination_expected();
45void rs_numberCountingCombination_observed();
46void rs_numberCountingCombination_observedWithTau();
47
48// -------------------------------
49// main driver to choose one
50void rs_numberCountingCombination(int flag = 1)
51{
52 if (flag == 1)
53 rs_numberCountingCombination_expected();
54 if (flag == 2)
55 rs_numberCountingCombination_observed();
56 if (flag == 3)
57 rs_numberCountingCombination_observedWithTau();
58}
59
60// -------------------------------
61void rs_numberCountingCombination_expected()
62{
63
64 /////////////////////////////////////////
65 // An example of a number counting combination with two channels.
66 // We consider both hypothesis testing and the equivalent confidence interval.
67 /////////////////////////////////////////
68
69 /////////////////////////////////////////
70 // The Model building stage
71 /////////////////////////////////////////
72
73 // Step 1, define arrays with signal & bkg expectations and background uncertainties
74 Double_t s[2] = {20., 10.}; // expected signal
75 Double_t b[2] = {100., 100.}; // expected background
76 Double_t db[2] = {.0100, .0100}; // fractional background uncertainty
77
78 // Step 2, use a RooStats factory to build a PDF for a
79 // number counting combination and add it to the workspace.
80 // We need to give the signal expectation to relate the masterSignal
81 // to the signal contribution in the individual channels.
82 // The model neglects correlations in background uncertainty,
83 // but they could be added without much change to the example.
85 RooWorkspace *wspace = new RooWorkspace();
86 f.AddModel(s, 2, wspace, "TopLevelPdf", "masterSignal");
87
88 // Step 3, use a RooStats factory to add datasets to the workspace.
89 // Step 3a.
90 // Add the expected data to the workspace
91 f.AddExpData(s, b, db, 2, wspace, "ExpectedNumberCountingData");
92
93 // see below for a printout of the workspace
94 // wspace->Print(); //uncomment to see structure of workspace
95
96 /////////////////////////////////////////
97 // The Hypothesis testing stage:
98 /////////////////////////////////////////
99 // Step 4, Define the null hypothesis for the calculator
100 // Here you need to know the name of the variables corresponding to hypothesis.
101 RooRealVar *mu = wspace->var("masterSignal");
102 RooArgSet *poi = new RooArgSet(*mu);
103 RooArgSet *nullParams = new RooArgSet("nullParams");
104 nullParams->addClone(*mu);
105 // here we explicitly set the value of the parameters for the null
106 nullParams->setRealValue("masterSignal", 0);
107
108 // Step 5, Create a calculator for doing the hypothesis test.
109 // because this is a
110 ProfileLikelihoodCalculator plc(*wspace->data("ExpectedNumberCountingData"), *wspace->pdf("TopLevelPdf"), *poi, 0.05,
111 nullParams);
112
113 // Step 6, Use the Calculator to get a HypoTestResult
114 HypoTestResult *htr = plc.GetHypoTest();
115 assert(htr != 0);
116 cout << "-------------------------------------------------" << endl;
117 cout << "The p-value for the null is " << htr->NullPValue() << endl;
118 cout << "Corresponding to a significance of " << htr->Significance() << endl;
119 cout << "-------------------------------------------------\n\n" << endl;
120
121 /* expected case should return:
122 -------------------------------------------------
123 The p-value for the null is 0.015294
124 Corresponding to a significance of 2.16239
125 -------------------------------------------------
126 */
127
128 //////////////////////////////////////////
129 // Confidence Interval Stage
130
131 // Step 8, Here we re-use the ProfileLikelihoodCalculator to return a confidence interval.
132 // We need to specify what are our parameters of interest
133 RooArgSet *paramsOfInterest = nullParams; // they are the same as before in this case
134 plc.SetParameters(*paramsOfInterest);
135 LikelihoodInterval *lrint = (LikelihoodInterval *)plc.GetInterval(); // that was easy.
136 lrint->SetConfidenceLevel(0.95);
137
138 // Step 9, make a plot of the likelihood ratio and the interval obtained
139 // paramsOfInterest->setRealValue("masterSignal",1.);
140 // find limits
141 double lower = lrint->LowerLimit(*mu);
142 double upper = lrint->UpperLimit(*mu);
143
144 LikelihoodIntervalPlot lrPlot(lrint);
145 lrPlot.SetMaximum(3.);
146 lrPlot.Draw();
147
148 // Step 10a. Get upper and lower limits
149 cout << "lower limit on master signal = " << lower << endl;
150 cout << "upper limit on master signal = " << upper << endl;
151
152 // Step 10b, Ask if masterSignal=0 is in the interval.
153 // Note, this is equivalent to the question of a 2-sigma hypothesis test:
154 // "is the parameter point masterSignal=0 inside the 95% confidence interval?"
155 // Since the significance of the Hypothesis test was > 2-sigma it should not be:
156 // eg. we exclude masterSignal=0 at 95% confidence.
157 paramsOfInterest->setRealValue("masterSignal", 0.);
158 cout << "-------------------------------------------------" << endl;
159 std::cout << "Consider this parameter point:" << std::endl;
160 paramsOfInterest->first()->Print();
161 if (lrint->IsInInterval(*paramsOfInterest))
162 std::cout << "It IS in the interval." << std::endl;
163 else
164 std::cout << "It is NOT in the interval." << std::endl;
165 cout << "-------------------------------------------------\n\n" << endl;
166
167 // Step 10c, We also ask about the parameter point masterSignal=2, which is inside the interval.
168 paramsOfInterest->setRealValue("masterSignal", 2.);
169 cout << "-------------------------------------------------" << endl;
170 std::cout << "Consider this parameter point:" << std::endl;
171 paramsOfInterest->first()->Print();
172 if (lrint->IsInInterval(*paramsOfInterest))
173 std::cout << "It IS in the interval." << std::endl;
174 else
175 std::cout << "It is NOT in the interval." << std::endl;
176 cout << "-------------------------------------------------\n\n" << endl;
177
178 delete lrint;
179 delete htr;
180 delete wspace;
181 delete poi;
182 delete nullParams;
183
184 /*
185 // Here's an example of what is in the workspace
186 // wspace->Print();
187 RooWorkspace(NumberCountingWS) Number Counting WS contents
188
189 variables
190 ---------
191 (x_0,masterSignal,expected_s_0,b_0,y_0,tau_0,x_1,expected_s_1,b_1,y_1,tau_1)
192
193 p.d.f.s
194 -------
195 RooProdPdf::joint[ pdfs=(sigRegion_0,sideband_0,sigRegion_1,sideband_1) ] = 2.20148e-08
196 RooPoisson::sigRegion_0[ x=x_0 mean=splusb_0 ] = 0.036393
197 RooPoisson::sideband_0[ x=y_0 mean=bTau_0 ] = 0.00398939
198 RooPoisson::sigRegion_1[ x=x_1 mean=splusb_1 ] = 0.0380088
199 RooPoisson::sideband_1[ x=y_1 mean=bTau_1 ] = 0.00398939
200
201 functions
202 --------
203 RooAddition::splusb_0[ set1=(s_0,b_0) set2=() ] = 120
204 RooProduct::s_0[ compRSet=(masterSignal,expected_s_0) compCSet=() ] = 20
205 RooProduct::bTau_0[ compRSet=(b_0,tau_0) compCSet=() ] = 10000
206 RooAddition::splusb_1[ set1=(s_1,b_1) set2=() ] = 110
207 RooProduct::s_1[ compRSet=(masterSignal,expected_s_1) compCSet=() ] = 10
208 RooProduct::bTau_1[ compRSet=(b_1,tau_1) compCSet=() ] = 10000
209
210 datasets
211 --------
212 RooDataSet::ExpectedNumberCountingData(x_0,y_0,x_1,y_1)
213
214 embedded pre-calculated expensive components
215 -------------------------------------------
216 */
217}
218
219void rs_numberCountingCombination_observed()
220{
221
222 /////////////////////////////////////////
223 // The same example with observed data in a main
224 // measurement and an background-only auxiliary
225 // measurement with a factor tau more background
226 // than in the main measurement.
227
228 /////////////////////////////////////////
229 // The Model building stage
230 /////////////////////////////////////////
231
232 // Step 1, define arrays with signal & bkg expectations and background uncertainties
233 // We still need the expectation to relate signal in different channels with the master signal
234 Double_t s[2] = {20., 10.}; // expected signal
235
236 // Step 2, use a RooStats factory to build a PDF for a
237 // number counting combination and add it to the workspace.
238 // We need to give the signal expectation to relate the masterSignal
239 // to the signal contribution in the individual channels.
240 // The model neglects correlations in background uncertainty,
241 // but they could be added without much change to the example.
243 RooWorkspace *wspace = new RooWorkspace();
244 f.AddModel(s, 2, wspace, "TopLevelPdf", "masterSignal");
245
246 // Step 3, use a RooStats factory to add datasets to the workspace.
247 // Add the observed data to the workspace
248 Double_t mainMeas[2] = {123., 117.}; // observed main measurement
249 Double_t bkgMeas[2] = {111.23, 98.76}; // observed background
250 Double_t dbMeas[2] = {.011, .0095}; // observed fractional background uncertainty
251 f.AddData(mainMeas, bkgMeas, dbMeas, 2, wspace, "ObservedNumberCountingData");
252
253 // see below for a printout of the workspace
254 // wspace->Print(); //uncomment to see structure of workspace
255
256 /////////////////////////////////////////
257 // The Hypothesis testing stage:
258 /////////////////////////////////////////
259 // Step 4, Define the null hypothesis for the calculator
260 // Here you need to know the name of the variables corresponding to hypothesis.
261 RooRealVar *mu = wspace->var("masterSignal");
262 RooArgSet *poi = new RooArgSet(*mu);
263 RooArgSet *nullParams = new RooArgSet("nullParams");
264 nullParams->addClone(*mu);
265 // here we explicitly set the value of the parameters for the null
266 nullParams->setRealValue("masterSignal", 0);
267
268 // Step 5, Create a calculator for doing the hypothesis test.
269 // because this is a
270 ProfileLikelihoodCalculator plc(*wspace->data("ObservedNumberCountingData"), *wspace->pdf("TopLevelPdf"), *poi, 0.05,
271 nullParams);
272
273 wspace->var("tau_0")->Print();
274 wspace->var("tau_1")->Print();
275
276 // Step 7, Use the Calculator to get a HypoTestResult
277 HypoTestResult *htr = plc.GetHypoTest();
278 cout << "-------------------------------------------------" << endl;
279 cout << "The p-value for the null is " << htr->NullPValue() << endl;
280 cout << "Corresponding to a significance of " << htr->Significance() << endl;
281 cout << "-------------------------------------------------\n\n" << endl;
282
283 /* observed case should return:
284 -------------------------------------------------
285 The p-value for the null is 0.0351669
286 Corresponding to a significance of 1.80975
287 -------------------------------------------------
288 */
289
290 //////////////////////////////////////////
291 // Confidence Interval Stage
292
293 // Step 8, Here we re-use the ProfileLikelihoodCalculator to return a confidence interval.
294 // We need to specify what are our parameters of interest
295 RooArgSet *paramsOfInterest = nullParams; // they are the same as before in this case
296 plc.SetParameters(*paramsOfInterest);
297 LikelihoodInterval *lrint = (LikelihoodInterval *)plc.GetInterval(); // that was easy.
298 lrint->SetConfidenceLevel(0.95);
299
300 // Step 9c. Get upper and lower limits
301 cout << "lower limit on master signal = " << lrint->LowerLimit(*mu) << endl;
302 cout << "upper limit on master signal = " << lrint->UpperLimit(*mu) << endl;
303
304 delete lrint;
305 delete htr;
306 delete wspace;
307 delete nullParams;
308 delete poi;
309}
310
311void rs_numberCountingCombination_observedWithTau()
312{
313
314 /////////////////////////////////////////
315 // The same example with observed data in a main
316 // measurement and an background-only auxiliary
317 // measurement with a factor tau more background
318 // than in the main measurement.
319
320 /////////////////////////////////////////
321 // The Model building stage
322 /////////////////////////////////////////
323
324 // Step 1, define arrays with signal & bkg expectations and background uncertainties
325 // We still need the expectation to relate signal in different channels with the master signal
326 Double_t s[2] = {20., 10.}; // expected signal
327
328 // Step 2, use a RooStats factory to build a PDF for a
329 // number counting combination and add it to the workspace.
330 // We need to give the signal expectation to relate the masterSignal
331 // to the signal contribution in the individual channels.
332 // The model neglects correlations in background uncertainty,
333 // but they could be added without much change to the example.
335 RooWorkspace *wspace = new RooWorkspace();
336 f.AddModel(s, 2, wspace, "TopLevelPdf", "masterSignal");
337
338 // Step 3, use a RooStats factory to add datasets to the workspace.
339 // Add the observed data to the workspace in the on-off problem.
340 Double_t mainMeas[2] = {123., 117.}; // observed main measurement
341 Double_t sideband[2] = {11123., 9876.}; // observed sideband
342 Double_t tau[2] = {100., 100.}; // ratio of bkg in sideband to bkg in main measurement, from experimental design.
343 f.AddDataWithSideband(mainMeas, sideband, tau, 2, wspace, "ObservedNumberCountingDataWithSideband");
344
345 // see below for a printout of the workspace
346 // wspace->Print(); //uncomment to see structure of workspace
347
348 /////////////////////////////////////////
349 // The Hypothesis testing stage:
350 /////////////////////////////////////////
351 // Step 4, Define the null hypothesis for the calculator
352 // Here you need to know the name of the variables corresponding to hypothesis.
353 RooRealVar *mu = wspace->var("masterSignal");
354 RooArgSet *poi = new RooArgSet(*mu);
355 RooArgSet *nullParams = new RooArgSet("nullParams");
356 nullParams->addClone(*mu);
357 // here we explicitly set the value of the parameters for the null
358 nullParams->setRealValue("masterSignal", 0);
359
360 // Step 5, Create a calculator for doing the hypothesis test.
361 // because this is a
362 ProfileLikelihoodCalculator plc(*wspace->data("ObservedNumberCountingDataWithSideband"), *wspace->pdf("TopLevelPdf"),
363 *poi, 0.05, nullParams);
364
365 // Step 7, Use the Calculator to get a HypoTestResult
366 HypoTestResult *htr = plc.GetHypoTest();
367 cout << "-------------------------------------------------" << endl;
368 cout << "The p-value for the null is " << htr->NullPValue() << endl;
369 cout << "Corresponding to a significance of " << htr->Significance() << endl;
370 cout << "-------------------------------------------------\n\n" << endl;
371
372 /* observed case should return:
373 -------------------------------------------------
374 The p-value for the null is 0.0352035
375 Corresponding to a significance of 1.80928
376 -------------------------------------------------
377 */
378
379 //////////////////////////////////////////
380 // Confidence Interval Stage
381
382 // Step 8, Here we re-use the ProfileLikelihoodCalculator to return a confidence interval.
383 // We need to specify what are our parameters of interest
384 RooArgSet *paramsOfInterest = nullParams; // they are the same as before in this case
385 plc.SetParameters(*paramsOfInterest);
386 LikelihoodInterval *lrint = (LikelihoodInterval *)plc.GetInterval(); // that was easy.
387 lrint->SetConfidenceLevel(0.95);
388
389 // Step 9c. Get upper and lower limits
390 cout << "lower limit on master signal = " << lrint->LowerLimit(*mu) << endl;
391 cout << "upper limit on master signal = " << lrint->UpperLimit(*mu) << endl;
392
393 delete lrint;
394 delete htr;
395 delete wspace;
396 delete nullParams;
397 delete poi;
398}
#define b(i)
Definition: RSha256.hxx:100
#define f(i)
Definition: RSha256.hxx:104
double Double_t
Definition: RtypesCore.h:55
virtual void Print(Option_t *options=0) const
Print TNamed name and title.
Definition: RooAbsArg.h:272
RooAbsArg * first() const
RooArgSet is a container object that can hold multiple RooAbsArg objects.
Definition: RooArgSet.h:28
Bool_t setRealValue(const char *name, Double_t newVal=0, Bool_t verbose=kFALSE)
Set value of a RooAbsRealLValye stored in set with given name to newVal No error messages are printed...
Definition: RooArgSet.cxx:493
virtual void addClone(const RooAbsCollection &col, Bool_t silent=kFALSE)
Add a collection of arguments to this collection by calling addOwned() for each element in the source...
Definition: RooArgSet.h:96
RooRealVar represents a fundamental (non-derived) real valued object.
Definition: RooRealVar.h:36
HypoTestResult is a base class for results from hypothesis tests.
virtual Double_t Significance() const
familiar name for the Null p-value in terms of 1-sided Gaussian significance
virtual Double_t NullPValue() const
Return p-value for null hypothesis.
This class provides simple and straightforward utilities to plot a LikelihoodInterval object.
LikelihoodInterval is a concrete implementation of the RooStats::ConfInterval interface.
Double_t LowerLimit(const RooRealVar &param)
return the lower bound of the interval on a given parameter
virtual void SetConfidenceLevel(Double_t cl)
set the confidence level for the interval (e.g 0.682 for a 1-sigma interval)
Double_t UpperLimit(const RooRealVar &param)
return the upper bound of the interval on a given parameter
virtual Bool_t IsInInterval(const RooArgSet &) const
check if given point is in the interval
A factory for building PDFs and data for a number counting combination.
The ProfileLikelihoodCalculator is a concrete implementation of CombinedCalculator (the interface cla...
The RooWorkspace is a persistable container for RooFit projects.
Definition: RooWorkspace.h:43
RooAbsData * data(const char *name) const
Retrieve dataset (binned or unbinned) with given name. A null pointer is returned if not found.
RooRealVar * var(const char *name) const
Retrieve real-valued variable (RooRealVar) with given name. A null pointer is returned if not found.
RooAbsPdf * pdf(const char *name) const
Retrieve p.d.f (RooAbsPdf) with given name. A null pointer is returned if not found.
Template specialisation used in RooAbsArg:
Namespace for the RooStats classes.
Definition: Asimov.h:20
static constexpr double s