Logo ROOT  
Reference Guide
rf202_extendedmlfit.py
Go to the documentation of this file.
1## \file
2## \ingroup tutorial_roofit
3## \notebook
4##
5## Addition and convolution: setting up an extended maximum likelihood fit
6##
7## \macro_code
8##
9## \date February 2018
10## \authors Clemens Lange, Wouter Verkerke (C++ version)
11
12import ROOT
13
14# Set up component pdfs
15# ---------------------------------------
16
17# Declare observable x
18x = ROOT.RooRealVar("x", "x", 0, 10)
19
20# Create two Gaussian PDFs g1(x,mean1,sigma) anf g2(x,mean2,sigma) and
21# their parameters
22mean = ROOT.RooRealVar("mean", "mean of gaussians", 5)
23sigma1 = ROOT.RooRealVar("sigma1", "width of gaussians", 0.5)
24sigma2 = ROOT.RooRealVar("sigma2", "width of gaussians", 1)
25
26sig1 = ROOT.RooGaussian("sig1", "Signal component 1", x, mean, sigma1)
27sig2 = ROOT.RooGaussian("sig2", "Signal component 2", x, mean, sigma2)
28
29# Build Chebychev polynomial p.d.f.
30a0 = ROOT.RooRealVar("a0", "a0", 0.5, 0., 1.)
31a1 = ROOT.RooRealVar("a1", "a1", -0.2, 0., 1.)
32bkg = ROOT.RooChebychev("bkg", "Background", x, ROOT.RooArgList(a0, a1))
33
34# Sum the signal components into a composite signal p.d.f.
35sig1frac = ROOT.RooRealVar(
36 "sig1frac", "fraction of component 1 in signal", 0.8, 0., 1.)
37sig = ROOT.RooAddPdf(
38 "sig", "Signal", ROOT.RooArgList(sig1, sig2), ROOT.RooArgList(sig1frac))
39
40# Method 1 - Construct extended composite model
41# -------------------------------------------------------------------
42
43# Sum the composite signal and background into an extended pdf
44# nsig*sig+nbkg*bkg
45nsig = ROOT.RooRealVar("nsig", "number of signal events", 500, 0., 10000)
46nbkg = ROOT.RooRealVar(
47 "nbkg", "number of background events", 500, 0, 10000)
48model = ROOT.RooAddPdf(
49 "model",
50 "(g1+g2)+a",
51 ROOT.RooArgList(
52 bkg,
53 sig),
54 ROOT.RooArgList(
55 nbkg,
56 nsig))
57
58# Sample, fit and plot extended model
59# ---------------------------------------------------------------------
60
61# Generate a data sample of expected number events in x from model
62# = model.expectedEvents() = nsig+nbkg
63data = model.generate(ROOT.RooArgSet(x))
64
65# Fit model to data, ML term automatically included
66model.fitTo(data)
67
68# Plot data and PDF overlaid, expected number of events for p.d.f projection normalization
69# rather than observed number of events (==data.numEntries())
70xframe = x.frame(ROOT.RooFit.Title("extended ML fit example"))
71data.plotOn(xframe)
72model.plotOn(xframe, ROOT.RooFit.Normalization(
73 1.0, ROOT.RooAbsReal.RelativeExpected))
74
75# Overlay the background component of model with a dashed line
76ras_bkg = ROOT.RooArgSet(bkg)
77model.plotOn(
78 xframe, ROOT.RooFit.Components(ras_bkg), ROOT.RooFit.LineStyle(
79 ROOT.kDashed), ROOT.RooFit.Normalization(
80 1.0, ROOT.RooAbsReal.RelativeExpected))
81
82# Overlay the background+sig2 components of model with a dotted line
83ras_bkg_sig2 = ROOT.RooArgSet(bkg, sig2)
84model.plotOn(
85 xframe, ROOT.RooFit.Components(ras_bkg_sig2), ROOT.RooFit.LineStyle(
86 ROOT.kDotted), ROOT.RooFit.Normalization(
87 1.0, ROOT.RooAbsReal.RelativeExpected))
88
89# Print structure of composite p.d.f.
90model.Print("t")
91
92
93# Method 2 - Construct extended components first
94# ---------------------------------------------------------------------
95
96# Associated nsig/nbkg as expected number of events with sig/bkg
97esig = ROOT.RooExtendPdf("esig", "extended signal p.d.f", sig, nsig)
98ebkg = ROOT.RooExtendPdf("ebkg", "extended background p.d.f", bkg, nbkg)
99
100# Sum extended components without coefs
101# -------------------------------------------------------------------------
102
103# Construct sum of two extended p.d.f. (no coefficients required)
104model2 = ROOT.RooAddPdf("model2", "(g1+g2)+a", ROOT.RooArgList(ebkg, esig))
105
106# Draw the frame on the canvas
107c = ROOT.TCanvas("rf202_extendedmlfit", "rf202_extendedmlfit", 600, 600)
108ROOT.gPad.SetLeftMargin(0.15)
109xframe.GetYaxis().SetTitleOffset(1.4)
110xframe.Draw()
111
112c.SaveAs("rf202_extendedmlfit.png")