`scipy.stats._probability_distribution:_ProbabilityDistribution.sample`

source: build-install/usr/lib/python3.14/site-packages/scipy/stats/_probability_distribution.py:112

Signature

def sample(self, shape, *, method, rng)

Summary

Random sample from the distribution.

Parameters

shape : tuple of ints, default: ()

The shape of the sample to draw. If the parameters of the distribution underlying the random variable are arrays of shape param_shape, the output array will be of shape shape + param_shape.

method : {None, 'formula', 'inverse_transform'}

The strategy used to produce the sample. By default (None), the infrastructure chooses between the following options, listed in order of precedence.

'formula': an implementation specific to the distribution
'inverse_transform': generate a uniformly distributed sample and return the inverse CDF at these arguments.

Not all method options are available for all distributions. If the selected method is not available, a NotImplementedError' will be raised.

rng : `numpy.random.Generator` or `scipy.stats.qmc.QMCEngine`, optional

Pseudo- or quasi-random number generator state. When rng is None, a new numpy.random.Generator is created using entropy from the operating system. Types other than numpy.random.Generator and scipy.stats.qmc.QMCEngine are passed to numpy.random.default_rng to instantiate a Generator.

If rng is an instance of scipy.stats.qmc.QMCEngine configured to use scrambling and shape is not empty, then each slice along the zeroth axis of the result is a "quasi-independent", low-discrepancy sequence; that is, they are distinct sequences that can be treated as statistically independent for most practical purposes. Separate calls to sample produce new quasi-independent, low-discrepancy sequences. The dimensionality (d) of the provided QMCEngine is ignored; new instances of the same class and configuration options (scramble, optimization, and bits, where applicable) will be created with d=1.

Examples

Instantiate a distribution with the desired parameters:

import numpy as np
from scipy import stats
X = stats.Uniform(a=0., b=1.)

✓

Generate a pseudorandom sample:

x = X.sample((1000, 1))
octiles = (np.arange(8) + 1) / 8

✓

np.count_nonzero(x <= octiles, axis=0)

✗

array([ 134,  266,  383,  527,  641,  774,  888, 1000])

X = stats.Uniform(a=np.zeros((3, 1)), b=np.ones(2))

✓

X.a.shape,

✗

((3, 2),)

x = X.sample(shape=(5, 4))
x.shape

✓

(5, 4, 3, 2)

Aliases

scipy.stats._distribution_infrastructure._ProbabilityDistribution.sample