dists#

Test statistic distributions models.

Classes:

`TSD`(values[, n_zero])	Trials-based representation of a test statistic distribution.
`Chi2TSD`(values[, n_zero, threshold])	\(\chi^2\)-fit representation of a test statistic distribution.
`BinnedTSD`(values[, n_zero])	Histogram-based representation of a test statistic distribution.

Functions:

ts_to_p(bg, key, ts[, fit])

Convert TS to p-values, interpolating over the background grid.

class csky.dists.TSD(values, n_zero=0, **kw)[source]#

Bases: object

Trials-based representation of a test statistic distribution.

Several methods are in the spirit of scipy.stats distribution objects, but note that at least currently, the interface is not a perfect match.

Methods:

`__init__`(values[, n_zero])	Construct a Chi2TSD.
`sf`(x[, fit])	The survival function, similar to `scipy.stats` distributions.
`sf_nsigma`(x, **kw)	The survival function, but returning a value in terms of number of sigmas.
`cdf`(x)	The cumulative distribution function, similar to `scipy.stats` distributions.
`isf`(p[, fit])	The inverse survival function, similar to `scipy.stats` distributions.
`median`()	The median TS value.
`get_hist`(**kw)	Get a `histlite.Hist` representation of the distribution.

__init__(values, n_zero=0, **kw)[source]#

Construct a Chi2TSD.

Parameters:

values (array of float or utis.Arrays) – the test statistic values
n_zero (int) – number of trials not contained by values for which TS=0
kw (mapping) – additional arguments for scipy.stats.chi2.fit. By default, if loc and floc are not set, then floc=0 will be used; if scale and fscale are not set, then fscale=1 will be used.

sf(x, fit=False)[source]#: The survival function, similar to scipy.stats distributions.

sf_nsigma(x, **kw)[source]#: The survival function, but returning a value in terms of number of sigmas.

cdf(x)[source]#: The cumulative distribution function, similar to scipy.stats distributions.

isf(p, fit=False)[source]#: The inverse survival function, similar to scipy.stats distributions.

median()[source]#: The median TS value.

get_hist(**kw)[source]#: Get a histlite.Hist representation of the distribution.

class csky.dists.Chi2TSD(values, n_zero=0, threshold=0, **kw)[source]#

Bases: TSD

\(\chi^2\)-fit representation of a test statistic distribution.

Information about trials is retained, and a chi2 fit is applied to the region TS > 0.

Methods:

`__init__`(values[, n_zero, threshold])	Construct a Chi2TSD.
`sf`(x[, fit])	The survival function, similar to `scipy.stats` distributions.
`cdf`(x[, fit])	The cumulative distribution function, similar to `scipy.stats` distributions.
`isf`(p[, fit])	The inverse survival function, similar to `scipy.stats` distributions.
`median`([fit])	The median TS value.
`get_hist`(**kw)	Get a `histlite.Hist` representation of the distribution.
`sf_nsigma`(x, **kw)	The survival function, but returning a value in terms of number of sigmas.

Attributes:

chi2

The fitted scipy.stats.chi2 object.

__init__(values, n_zero=0, threshold=0, **kw)[source]#

Construct a Chi2TSD.

Parameters:

values (array of float or utis.Arrays) – the test statistic values
n_zero (int) – number of trials not contained by values for which TS=0
kw (mapping) – additional arguments for scipy.stats.chi2.fit. By default, if loc and floc are not set, then floc=0 will be used; if scale and fscale are not set, then fscale=1 will be used.

property chi2#: The fitted scipy.stats.chi2 object.

sf(x, fit=True)[source]#: The survival function, similar to scipy.stats distributions.

cdf(x, fit=True)[source]#: The cumulative distribution function, similar to scipy.stats distributions.

isf(p, fit=True)[source]#: The inverse survival function, similar to scipy.stats distributions.

median(fit=False)[source]#: The median TS value.

get_hist(**kw)#: Get a histlite.Hist representation of the distribution.

sf_nsigma(x, **kw)#: The survival function, but returning a value in terms of number of sigmas.

class csky.dists.BinnedTSD(values, n_zero=0, **kw)[source]#

Bases: TSD

Histogram-based representation of a test statistic distribution.

This class tries to reproduce the behavior of TSD without storing information about every single trial, thereby reducing RAM and disk space requirements.

Methods:

`__init__`(values[, n_zero])	Construct a Chi2TSD.
`median`()	The median TS value.
`sf`(x[, fit])	The survival function, similar to `scipy.stats` distributions.
`cdf`(x[, fit])	The cumulative distribution function, similar to `scipy.stats` distributions.
`get_hist`(**kw)	Get a `histlite.Hist` representation of the distribution.
`isf`(p[, fit])	The inverse survival function, similar to `scipy.stats` distributions.
`sf_nsigma`(x, **kw)	The survival function, but returning a value in terms of number of sigmas.

__init__(values, n_zero=0, **kw)[source]#

Construct a Chi2TSD.

Parameters:

values (array of float or utis.Arrays) – the test statistic values
n_zero (int) – number of trials not contained by values for which TS=0
kw (mapping) – additional arguments for scipy.stats.chi2.fit. By default, if loc and floc are not set, then floc=0 will be used; if scale and fscale are not set, then fscale=1 will be used.

median()[source]#: The median TS value.

sf(x, fit=True)[source]#: The survival function, similar to scipy.stats distributions.

cdf(x, fit=True)[source]#: The cumulative distribution function, similar to scipy.stats distributions.

get_hist(**kw)#: Get a histlite.Hist representation of the distribution.

isf(p, fit=True)[source]#: The inverse survival function, similar to scipy.stats distributions.

sf_nsigma(x, **kw)#: The survival function, but returning a value in terms of number of sigmas.

csky.dists.ts_to_p(bg, key, ts, fit=True)[source]#

Convert TS to p-values, interpolating over the background grid.

Parameters:

bg (dict) – the background distributions
key (array of float) – the values that indicate where along the background parameterization each TS value originates from
fit (bool) – whether to request sf(ts_value, fit=True)