knuth_bin_count#

scikitplot.feature_selection.knuth_bin_count(values, min_bins=1, max_bins=100)[source]#

Return the equal-width bin count that maximises Knuth’s posterior.

Parameters:

valuesarray-like of float: Finite sample values.
min_bins, max_binsint, default=1, 100: Inclusive search range.

Returns:

int: The count with the largest log posterior; the smallest such count on an exact tie.

Raises:

ValueError: For an empty sample, non-finite values or an invalid range.

Parameters:

values (Any)
min_bins (int)
max_bins (int)

Return type:

int

Notes

The log posterior of M equal-width bins with counts n_k over N values is (Knuth 2006, eq. 31):

N log M + lgamma(M/2) - M lgamma(1/2) - lgamma(N + M/2)
    + sum_k lgamma(n_k + 1/2)

Examples

>>> import numpy as np
>>> from scikitplot.feature_selection import knuth_bin_count
>>> knuth_bin_count(np.random.default_rng(0).normal(size=1000), 1, 50)
9