asv._stats

Attributes

_mann_whitney_u_memo

Functions

get_weight(stats)	Return a data point weight for the result.
is_different(samples_a, samples_b, stats_a, stats_b[, ...])	Check whether the samples are statistically different.
mann_whitney_u(x, y[, method])	Mann-Whitney U test
mann_whitney_u_u(x, y)
mann_whitney_u_cdf(m, n, u[, memo])
mann_whitney_u_pmf(m, n, u[, memo])
mann_whitney_u_r(m, n, u[, memo])	Number of orderings in Mann-Whitney U test.
binom(n, k)	Binomial coefficient (n over k)

Module Contents

asv._stats.get_weight(stats)

Return a data point weight for the result.

asv._stats.is_different(samples_a, samples_b, stats_a, stats_b, p_threshold=0.002)

Check whether the samples are statistically different.

If sample data is not provided, or the sample is too small, falls back to a pessimistic CI-based check. If it returns True, then the difference is statistically significant. If it returns False, it might or might not be statistically significant.

Parameters

samples_a, samples_b

Input samples

stats_a, stats_b

Input stats data

asv._stats._mann_whitney_u_memo

asv._stats.mann_whitney_u(x, y, method='auto')

Mann-Whitney U test

Ties are handled conservatively, returning the least significant tie breaking. .

Parameters

x, ylist of float

Samples to test

method{‘auto’, ‘exact’, ‘normal’}

Whether to compute p-value exactly of via normal approximation. The option ‘auto’ switches to approximation for sample size > 20.

Returns

uint

U-statistic

pfloat

p-value for two-sided alternative

References

[MWU_GC10]

Jean Dickinson Gibbons and Subhabrata Chakraborti. Nonparametric Statistical Inference. Chapman and Hall/CRC, New York, 5 edition, July 2010. ISBN 978-0-429-11188-4. doi:10.1201/9781439896129.

[MWU_MW47]

H. B. Mann and D. R. Whitney. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics, 18(1):50–60, 1947. arXiv:2236101, doi:10.1214/aoms/1177730491.

asv._stats.mann_whitney_u_u(x, y)

asv._stats.mann_whitney_u_cdf(m, n, u, memo=None)

asv._stats.mann_whitney_u_pmf(m, n, u, memo=None)

asv._stats.mann_whitney_u_r(m, n, u, memo=None)

Number of orderings in Mann-Whitney U test.

The PMF of U for samples of sizes (m, n) is given by

p(u) = r(m, n, u) / binom(m + n, m).

References

[MWUR_MW47]

H. B. Mann and D. R. Whitney. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics, 18(1):50–60, 1947. arXiv:2236101, doi:10.1214/aoms/1177730491.

asv._stats.binom(n, k)

Binomial coefficient (n over k)