Logo ROOT  
Reference Guide
goftest.C File Reference

Detailed Description

View in nbviewer Open in SWAN GoFTest tutorial macro

Using Anderson-Darling and Kolmogorov-Smirnov goodness of fit tests 1 sample test is performed comparing data with a log-normal distribution and a 2 sample test is done comparing two gaussian data sets.

TEST with LANDAU distribution: OK ( pvalues = 0.777278 )
#include <cassert>
#include "TCanvas.h"
#include "TPaveText.h"
#include "TH1.h"
#include "TF1.h"
#include "Math/GoFTest.h"
#include "Math/Functor.h"
#include "TRandom3.h"
#include "Math/DistFunc.h"
// need to use Functor1D
double landau(double x) {
}
void goftest() {
// ------------------------------------------------------------------------
// Case 1: Create logNormal random sample
UInt_t nEvents1 = 1000;
//ROOT::Math::Random<ROOT::Math::GSLRngMT> r;
TF1 * f1 = new TF1("logNormal","ROOT::Math::lognormal_pdf(x,[0],[1])",0,500);
// set the lognormal parameters (m and s)
f1->SetParameters(4.0,1.0);
f1->SetNpx(1000);
Double_t* sample1 = new Double_t[nEvents1];
TH1D* h1smp = new TH1D("h1smp", "LogNormal distribution histogram", 100, 0, 500);
h1smp->SetStats(kFALSE);
for (UInt_t i = 0; i < nEvents1; ++i) {
//Double_t data = f1->GetRandom();
Double_t data = gRandom->Gaus(4,1);
data = TMath::Exp(data);
sample1[i] = data;
h1smp->Fill(data);
}
// normalize correctly the histogram using the entries inside
h1smp->Scale( ROOT::Math::lognormal_cdf(500.,4.,1) / nEvents1, "width");
TCanvas* c = new TCanvas("c","1-Sample and 2-Samples GoF Tests");
c->Divide(1, 2);
TPad * pad = (TPad *)c->cd(1);
h1smp->Draw();
pad->SetLogy();
f1->SetNpx(100); // use same points as histo for drawing
f1->Draw("SAME");
// -----------------------------------------
// Create GoFTest object
//----------------------------------------------------
// Possible calls for the Anderson - DarlingTest test
// a) Returning the Anderson-Darling standardized test statistic
Double_t A2_1 = goftest_1-> AndersonDarlingTest("t");
Double_t A2_2 = (*goftest_1)(ROOT::Math::GoFTest::kAD, "t");
assert(A2_1 == A2_2);
// b) Returning the p-value for the Anderson-Darling test statistic
Double_t pvalueAD_1 = goftest_1-> AndersonDarlingTest(); // p-value is the default choice
Double_t pvalueAD_2 = (*goftest_1)(); // p-value and Anderson - Darling Test are the default choices
assert(pvalueAD_1 == pvalueAD_2);
// Rebuild the test using the default 1-sample construtor
delete goftest_1;
goftest_1 = new ROOT::Math::GoFTest(nEvents1, sample1 ); // User must then input a distribution type option
//--------------------------------------------------
// Possible calls for the Kolmogorov - Smirnov test
// a) Returning the Kolmogorov-Smirnov standardized test statistic
Double_t Dn_1 = goftest_1-> KolmogorovSmirnovTest("t");
Double_t Dn_2 = (*goftest_1)(ROOT::Math::GoFTest::kKS, "t");
assert(Dn_1 == Dn_2);
// b) Returning the p-value for the Kolmogorov-Smirnov test statistic
Double_t pvalueKS_1 = goftest_1-> KolmogorovSmirnovTest();
Double_t pvalueKS_2 = (*goftest_1)(ROOT::Math::GoFTest::kKS);
assert(pvalueKS_1 == pvalueKS_2);
// Valid but incorrect calls for the 2-samples methods of the 1-samples constructed goftest_1
#ifdef TEST_ERROR_MESSAGE
Double_t A2 = (*goftest_1)(ROOT::Math::GoFTest::kAD2s, "t"); // Issues error message
Double_t pvalueKS = (*goftest_1)(ROOT::Math::GoFTest::kKS2s); // Issues error message
assert(A2 == pvalueKS);
#endif
TPaveText* pt1 = new TPaveText(0.58, 0.6, 0.88, 0.80, "brNDC");
Char_t str1[50];
sprintf(str1, "p-value for A-D 1-smp test: %f", pvalueAD_1);
pt1->AddText(str1);
pt1->SetFillColor(18);
pt1->SetTextFont(20);
pt1->SetTextColor(4);
Char_t str2[50];
sprintf(str2, "p-value for K-S 1-smp test: %f", pvalueKS_1);
pt1->AddText(str2);
pt1->Draw();
// ------------------------------------------------------------------------
// Case 2: Create Gaussian random samples
UInt_t nEvents2 = 2000;
Double_t* sample2 = new Double_t[nEvents2];
TH1D* h2smps_1 = new TH1D("h2smps_1", "Gaussian distribution histograms", 100, 0, 500);
h2smps_1->SetStats(kFALSE);
TH1D* h2smps_2 = new TH1D("h2smps_2", "Gaussian distribution histograms", 100, 0, 500);
h2smps_2->SetStats(kFALSE);
for (UInt_t i = 0; i < nEvents1; ++i) {
Double_t data = r.Gaus(300, 50);
sample1[i] = data;
h2smps_1->Fill(data);
}
h2smps_1->Scale(1. / nEvents1, "width");
c->cd(2);
h2smps_1->Draw();
h2smps_1->SetLineColor(kBlue);
for (UInt_t i = 0; i < nEvents2; ++i) {
Double_t data = r.Gaus(300, 50);
sample2[i] = data;
h2smps_2->Fill(data);
}
h2smps_2->Scale(1. / nEvents2, "width");
h2smps_2->Draw("SAME");
h2smps_2->SetLineColor(kRed);
// -----------------------------------------
// Create GoFTest object
ROOT::Math::GoFTest* goftest_2 = new ROOT::Math::GoFTest(nEvents1, sample1, nEvents2, sample2);
//----------------------------------------------------
// Possible calls for the Anderson - DarlingTest test
// a) Returning the Anderson-Darling standardized test statistic
A2_1 = goftest_2->AndersonDarling2SamplesTest("t");
A2_2 = (*goftest_2)(ROOT::Math::GoFTest::kAD2s, "t");
assert(A2_1 == A2_2);
// b) Returning the p-value for the Anderson-Darling test statistic
pvalueAD_1 = goftest_2-> AndersonDarling2SamplesTest(); // p-value is the default choice
pvalueAD_2 = (*goftest_2)(ROOT::Math::GoFTest::kAD2s); // p-value is the default choices
assert(pvalueAD_1 == pvalueAD_2);
//--------------------------------------------------
// Possible calls for the Kolmogorov - Smirnov test
// a) Returning the Kolmogorov-Smirnov standardized test statistic
Dn_1 = goftest_2-> KolmogorovSmirnov2SamplesTest("t");
Dn_2 = (*goftest_2)(ROOT::Math::GoFTest::kKS2s, "t");
assert(Dn_1 == Dn_2);
// b) Returning the p-value for the Kolmogorov-Smirnov test statistic
pvalueKS_1 = goftest_2-> KolmogorovSmirnov2SamplesTest();
pvalueKS_2 = (*goftest_2)(ROOT::Math::GoFTest::kKS2s);
assert(pvalueKS_1 == pvalueKS_2);
#ifdef TEST_ERROR_MESSAGE
/* Valid but incorrect calls for the 1-sample methods of the 2-samples constucted goftest_2 */
A2 = (*goftest_2)(ROOT::Math::GoFTest::kAD, "t"); // Issues error message
pvalueKS = (*goftest_2)(ROOT::Math::GoFTest::kKS); // Issues error message
assert(A2 == pvalueKS);
#endif
TPaveText* pt2 = new TPaveText(0.13, 0.6, 0.43, 0.8, "brNDC");
sprintf(str1, "p-value for A-D 2-smps test: %f", pvalueAD_1);
pt2->AddText(str1);
pt2->SetFillColor(18);
pt2->SetTextFont(20);
pt2->SetTextColor(4);
sprintf(str2, "p-value for K-S 2-smps test: %f", pvalueKS_1);
pt2-> AddText(str2);
pt2-> Draw();
// ------------------------------------------------------------------------
// Case 3: Create Landau random sample
UInt_t nEvents3 = 1000;
Double_t* sample3 = new Double_t[nEvents3];
for (UInt_t i = 0; i < nEvents3; ++i) {
Double_t data = r.Landau();
sample3[i] = data;
}
// ------------------------------------------
// Create GoFTest objects
//
// Possible constructors for the user input distribution
// a) User input PDF
double minimum = 3*TMath::MinElement(nEvents3, sample3);
double maximum = 3*TMath::MaxElement(nEvents3, sample3);
ROOT::Math::GoFTest* goftest_3a = new ROOT::Math::GoFTest(nEvents3, sample3, f, ROOT::Math::GoFTest::kPDF, minimum,maximum); // need to specify am interval
// b) User input CDF
ROOT::Math::GoFTest* goftest_3b = new ROOT::Math::GoFTest(nEvents3, sample3, fI, ROOT::Math::GoFTest::kCDF,minimum,maximum);
// Returning the p-value for the Anderson-Darling test statistic
pvalueAD_1 = goftest_3a-> AndersonDarlingTest(); // p-value is the default choice
pvalueAD_2 = (*goftest_3b)(); // p-value and Anderson - Darling Test are the default choices
// Checking consistency between both tests
std::cout << " \n\nTEST with LANDAU distribution:\t";
if (TMath::Abs(pvalueAD_1 - pvalueAD_2) > 1.E-1 * pvalueAD_2) {
std::cout << "FAILED " << std::endl;
Error("goftest","Error in comparing testing using Landau and Landau CDF");
std::cerr << " pvalues are " << pvalueAD_1 << " " << pvalueAD_2 << std::endl;
}
else
std::cout << "OK ( pvalues = " << pvalueAD_2 << " )" << std::endl;
}
ROOT::R::TRInterface & r
Definition: Object.C:4
#define f(i)
Definition: RSha256.hxx:104
#define c(i)
Definition: RSha256.hxx:101
char Char_t
Definition: RtypesCore.h:31
unsigned int UInt_t
Definition: RtypesCore.h:44
const Bool_t kFALSE
Definition: RtypesCore.h:90
double Double_t
Definition: RtypesCore.h:57
@ kRed
Definition: Rtypes.h:64
@ kBlue
Definition: Rtypes.h:64
void Error(const char *location, const char *msgfmt,...)
R__EXTERN TRandom * gRandom
Definition: TRandom.h:62
Functor1D class for one-dimensional functions.
Definition: Functor.h:493
void SetDistribution(EDistribution dist)
Definition: GoFTest.cxx:123
void AndersonDarling2SamplesTest(Double_t &pvalue, Double_t &testStat) const
Definition: GoFTest.cxx:646
virtual void SetFillColor(Color_t fcolor)
Set the fill area color.
Definition: TAttFill.h:37
virtual void SetLineColor(Color_t lcolor)
Set the line color.
Definition: TAttLine.h:40
virtual void SetTextColor(Color_t tcolor=1)
Set the text color.
Definition: TAttText.h:43
virtual void SetTextFont(Font_t tfont=62)
Set the text font.
Definition: TAttText.h:45
The Canvas class.
Definition: TCanvas.h:27
1-Dim function class
Definition: TF1.h:210
virtual void SetNpx(Int_t npx=100)
Set the number of points used to draw the function.
Definition: TF1.cxx:3435
virtual void Draw(Option_t *option="")
Draw this function with its current attributes.
Definition: TF1.cxx:1320
virtual void SetParameters(const Double_t *params)
Definition: TF1.h:638
1-D histogram with a double per channel (see TH1 documentation)}
Definition: TH1.h:614
virtual Int_t Fill(Double_t x)
Increment bin with abscissa X by 1.
Definition: TH1.cxx:3275
virtual void Draw(Option_t *option="")
Draw this histogram with options.
Definition: TH1.cxx:2998
virtual void Scale(Double_t c1=1, Option_t *option="")
Multiply this histogram by a constant c1.
Definition: TH1.cxx:6246
virtual void SetStats(Bool_t stats=kTRUE)
Set statistics option on/off.
Definition: TH1.cxx:8446
The most important graphics class in the ROOT system.
Definition: TPad.h:29
virtual void SetLogy(Int_t value=1)
Set Lin/Log scale for Y.
Definition: TPad.cxx:5904
A Pave (see TPave) with text, lines or/and boxes inside.
Definition: TPaveText.h:21
virtual TText * AddText(Double_t x1, Double_t y1, const char *label)
Add a new Text line to this pavetext at given coordinates.
Definition: TPaveText.cxx:182
virtual void Draw(Option_t *option="")
Draw this pavetext with its current attributes.
Definition: TPaveText.cxx:233
Random number generator class based on M.
Definition: TRandom3.h:27
virtual Double_t Gaus(Double_t mean=0, Double_t sigma=1)
Samples a random number from the standard Normal (Gaussian) Distribution with the given mean and sigm...
Definition: TRandom.cxx:263
double landau_pdf(double x, double xi=1, double x0=0)
Probability density function of the Landau distribution:
double lognormal_cdf(double x, double m, double s, double x0=0)
Cumulative distribution function of the lognormal distribution (lower tail).
Double_t x[n]
Definition: legend1.C:17
TF1 * f1
Definition: legend1.C:11
Double_t Exp(Double_t x)
Definition: TMath.h:717
T MinElement(Long64_t n, const T *a)
Return minimum of array a of length n.
Definition: TMath.h:942
T MaxElement(Long64_t n, const T *a)
Return maximum of array a of length n.
Definition: TMath.h:949
Double_t LandauI(Double_t x)
Returns the value of the Landau distribution function at point x.
Definition: TMath.cxx:2796
Short_t Abs(Short_t d)
Definition: TMathBase.h:120
th1 Draw()
Author
Bartolomeu Rabacal

Definition in file goftest.C.