sensitivity.cpp File Reference

Calculate sensitivity with an MCMC. More...

#include <iostream>
#include <string>
#include <algorithm>
#include <assert.h>
#include <TFile.h>
#include <TH1F.h>
#include <TNtuple.h>
#include <TCanvas.h>
#include <TRandom.h>
#include <TRandom2.h>
#include "config.h"
#include "generator.h"
#include "mcmc.h"

Functions
double	upper_limit (TH1F *h, double cl=0.682)
TH1F *	ensemble (std::vector< Signal > &signals, Range< float > &e_range, Range< float > &r_range, unsigned steps, float burnin_fraction, std::string signal_name, float confidence, unsigned nexperiments)
int	main (int argc, char *argv[])

---

Detailed Description

Calculate sensitivity with an MCMC.

An ensemble of fake experiments is generated and each fit with an MCMC, with parameters defined in the JSON configuration file. Limits are calculated for each.

---

Function Documentation

TH1F* ensemble	(	std::vector< Signal > &	signals,
		Range< float > &	e_range,
		Range< float > &	r_range,
		unsigned	steps,
		float	burnin_fraction,
		std::string	signal_name,
		float	confidence,
		unsigned	nexperiments
	)

Run an ensemble of independent fake experiments

Run experiments, fit with mcmc, and tabulate background-fluctuation sensitivity for each, creating a histogram of limits. The estimated sensitivity is the median of the limits of the ensemble.

Parameters:

	signals	List of Signals defining PDFs, rates, etc.
	e_range	Energy range to consider
	r_range	Radial range to consider
	steps	Number of MCMC random walk steps to take
	burnin_fraction	Fraction of initial MCMC steps to throw out
	signal_name	The name of the Signal that is the signal
	confidence	Desired confidence level for limits
	nexperiments	Number of fake experiments to run

Returns:

A histogram of limits from the experiments

int main	(	int	argc,
		char *	argv[]
	)

Run the sensitivity calculation

Parameters:

	argc	Argument count: must be 2
	argv	Arguments: executable path and configuration filename

Returns:

0 on success, non-zero on failure

double upper_limit	(	TH1F *	h,
		double	cl = 0.682
	)

Find an upper limit

Locate the x value that a fraction 1-CL/2 of the distribution falls above. This may over-cover due to finite histogram bin widths.

Parameters:

	h	1D histogram to calculate limit on
	cl	Confidence level

Returns:

The nearest bin edge covering at least the desired confidence