stats - Statistics

Introduction

gammapy.stats holds statistical estimators, fit statistics and algorithms commonly used in gamma-ray astronomy.

It is mostly concerned with the evaluation of one or several observations that count events in a given region and time window, i.e. with Poisson-distributed counts measurements.

For on-off methods we will use the following variable names following the notation in [Cousins2007]:

Variable Definition
n_on Total observed counts in the on region
n_off Total observed counts in the off region
mu_on Total expected counts in the on region
mu_off Total expected counts in the off region
mu_sig Signal expected counts in the on region
mu_bkg Background expected counts in the on region
a_on Relative background exposure in the on region
a_off Relative background exposure in the off region
alpha Background efficiency ratio a_on / a_off
n_bkg Background estimate in the on region

The following formulae show how an on-off measurement \((n_{on}, n_{off})\) is related to the quantities in the above table:

\[n_{on} \sim Pois(\mu_{on})\text{ with }\mu_{on} = \mu_s + \mu_b n_{off} \sim Pois(\mu_{off})\text{ with }\mu_{off} = \mu_b / \alpha\text{ with }\alpha = a_{on} / a_{off}\]

With the background estimate in the on region

\[n_{bkg} = \alpha\ n_{off},\]

the maximum likelihood estimate of a signal excess is

\[n_{excess} = n_{on} - n_{bkg}.\]

When the background is known and there is only an “on” region (sometimes also called “source region”), we use the variable names n_on, mu_on, mu_sig and mu_bkg.

These are references describing the available methods: [LiMa1983], [Cash1979], [Stewart2009], [Rolke2005], [Feldman1998], [Cousins2007].

Getting Started

Li & Ma Significance

[LiMa1983] (see equation 17)

As an example, assume you measured \(n_{on} = 18\) counts in a region where you suspect a source might be present and \(n_{off} = 97\) counts in a background control region where you assume no source is present and that is \(a_{off}/a_{on}=10\) times larger than the on-region.

Here’s how you compute the statistical significance of your detection with the Li & Ma formula:

>>> from gammapy.stats import significance_on_off
>>> significance_on_off(n_on=18, n_off=97, alpha=1. / 10, method='lima')
2.2421704424844875

Confidence Intervals

Assume you measured 6 counts in a Poissonian counting experiment with an expected background \(b = 3\). Here’s how you compute the 90% upper limit on the signal strength \(\\mu\):

import numpy as np
from scipy import stats
import gammapy.stats as gstats

x_bins = np.arange(0, 100)
mu_bins = np.linspace(0, 50, 50 / 0.005 + 1, endpoint=True)

matrix = [stats.poisson(mu + 3).pmf(x_bins) for mu in mu_bins]
acceptance_intervals = gstats.fc_construct_acceptance_intervals_pdfs(matrix, 0.9)
LowerLimitNum, UpperLimitNum, _ = gstats.fc_get_limits(mu_bins, x_bins, acceptance_intervals)
mu_upper_limit = gstats.fc_find_limit(6, UpperLimitNum, mu_bins)

The result is mu_upper_limit == 8.465.

Reference/API

gammapy.stats Package

Statistics.

Functions

background(n_off, alpha) Estimate background in the on-region from an off-region observation.
background_error(n_off, alpha) Estimate standard error on background in the on region from an off-region observation.
cash(n_on, mu_on) Cash statistic, for Poisson data.
cash_sum_cython() Summed cash fit statistics.
combine_stats(stats_1, stats_2[, weight_method]) Combine using some weight method for the exposure.
compute_total_stats(counts, exposure[, …]) Compute total stats for arrays of per-bin stats.
cstat(n_on, mu_on[, n_on_min]) C statistic, for Poisson data.
cstat_sum_cython() Summed cstat fit statistics.
excess(n_on, n_off, alpha) Estimate excess in the on region for an on-off observation.
excess_error(n_on, n_off, alpha) Estimate error on excess for an on-off measurement.
excess_matching_significance(mu_bkg, …[, …]) Compute excess matching a given significance.
excess_matching_significance_on_off(n_off, …) Compute sensitivity of an on-off observation.
excess_ul_helene(excess, excess_error, …) Compute excess upper limit using the Helene method.
fc_construct_acceptance_intervals(…) Convenience function that calculates the PDF for the user.
fc_construct_acceptance_intervals_pdfs(…) Numerically choose bins a la Feldman Cousins ordering principle.
fc_find_acceptance_interval_gauss(mu, sigma, …) Analytical acceptance interval for Gaussian with boundary at the origin.
fc_find_acceptance_interval_poisson(mu, …) Analytical acceptance interval for Poisson process with background.
fc_find_average_upper_limit(x_bins, matrix, …) Function to calculate the average upper limit for a confidence belt
fc_find_limit(x_value, x_values, y_values) Find the limit for a given x measurement
fc_fix_limits(lower_limit, upper_limit) Push limits outwards as described in the FC paper.
fc_get_limits(mu_bins, x_bins, …) Find lower and upper limit from acceptance intervals.
get_wstat_gof_terms(n_on, n_off) Calculate goodness of fit terms for wstat
get_wstat_mu_bkg(n_on, n_off, alpha, mu_sig) Calculate mu_bkg for wstat
make_stats(signal, background, area_factor) Fill using some weight method for the exposure.
probability_to_significance_normal(probability) Convert one-sided tail probability to significance.
probability_to_significance_normal_limit(…) Convert tail probability to significance in the limit of small p and large s.
significance(n_on, mu_bkg[, method, n_on_min]) Compute significance for an observed number of counts and known background.
significance_on_off(n_on, n_off, alpha[, method]) Compute significance of an on-off observation.
significance_to_probability_normal(significance) Convert significance to one-sided tail probability.
significance_to_probability_normal_limit(…) Convert significance to tail probability in the limit of small p and large s.
wstat(n_on, n_off, alpha, mu_sig[, mu_bkg, …]) W statistic, for Poisson data with Poisson background.

Classes

Stats(n_on, n_off, a_on, a_off) Container for an on-off observation.