km3pipe.stats

Statistics.

Module Contents

Classes

loguniform(low=0.1, high=1, base=10, *args, **kwargs)	Loguniform Distributon
rv_kde(data, bw=None, bw_method=None, bw_statsmodels=False, **kde_args)	Create a scipy.stats.rv_continuous instance from a (gaussian) KDE.
hist2d(H2D, *args, **kwargs)	Simple implementation of a 2d histogram.

Functions

mad(v)	MAD – Median absolute deviation. More robust than standard deviation.
mad_std(v)	Robust estimate of standard deviation using the MAD.
drop_zero_variance(df)	Remove columns from dataframe with zero variance.
param_names(scipy_dist)	Get names of fit parameters from a scipy.rv_* distribution.
perc(arr, p=95, **kwargs)	Create symmetric percentiles, with p coverage.
resample_1d(arr, n_out=None, random_state=None)	Resample an array, with replacement.
bootstrap_params(rv_cont, data, n_iter=5, **kwargs)	Bootstrap the fit params of a distribution.
param_describe(params, quant=95, axis=0)	Get mean + quantile range from bootstrapped params.
bootstrap_fit(rv_cont, data, n_iter=10, quant=95, print_params=True, **kwargs)	Bootstrap a distribution fit + get confidence intervals for the params.
bincenters(bins)	Bincenters, assuming they are all equally spaced.

km3pipe.stats.log

class km3pipe.stats.loguniform(low=0.1, high=1, base=10, *args, **kwargs)

Bases:scipy.stats.rv_continuous

Loguniform Distributon

class km3pipe.stats.rv_kde(data, bw=None, bw_method=None, bw_statsmodels=False, **kde_args)

Bases:scipy.stats.rv_continuous

Create a scipy.stats.rv_continuous instance from a (gaussian) KDE.

Uses the KDE implementation from sklearn.

Automatic bandwidth, either from the statsmodels or scipy implementation.

pdf(self, x)

rvs(self, *args, **kwargs)

Draw Random Variates.

Parameters:

size: int, optional (default=1)

random_state_: optional (default=None)

km3pipe.stats.mad(v)

MAD – Median absolute deviation. More robust than standard deviation.

km3pipe.stats.mad_std(v)

Robust estimate of standard deviation using the MAD.

Lifted from astropy.stats.

km3pipe.stats.drop_zero_variance(df)

Remove columns from dataframe with zero variance.

km3pipe.stats.param_names(scipy_dist)

Get names of fit parameters from a scipy.rv_* distribution.

km3pipe.stats.perc(arr, p=95, **kwargs)

Create symmetric percentiles, with p coverage.

km3pipe.stats.resample_1d(arr, n_out=None, random_state=None)

Resample an array, with replacement.

Parameters:

arr: np.ndarray

The array is resampled along the first axis.

n_out: int, optional

Number of samples to return. If not specified, return len(arr) samples.

km3pipe.stats.bootstrap_params(rv_cont, data, n_iter=5, **kwargs)

Bootstrap the fit params of a distribution.

Parameters:

rv_cont: scipy.stats.rv_continuous instance

The distribution which to fit.

data: array-like, 1d

The data on which to fit.

n_iter: int [default=10]

Number of bootstrap iterations.

km3pipe.stats.param_describe(params, quant=95, axis=0)

Get mean + quantile range from bootstrapped params.

km3pipe.stats.bootstrap_fit(rv_cont, data, n_iter=10, quant=95, print_params=True, **kwargs)

Bootstrap a distribution fit + get confidence intervals for the params.

Parameters:

rv_cont: scipy.stats.rv_continuous instance

The distribution which to fit.

data: array-like, 1d

The data on which to fit.

n_iter: int [default=10]

Number of bootstrap iterations.

quant: int [default=95]

percentile of the confidence limits (default is 95, i.e. 2.5%-97.5%)

print_params: bool [default=True]

Print a fit summary.

class km3pipe.stats.hist2d(H2D, *args, **kwargs)

Bases:scipy.stats.rv_continuous

Simple implementation of a 2d histogram.

H_pad

integral

area

km3pipe.stats.bincenters(bins)

Bincenters, assuming they are all equally spaced.