Skip to main content

Module statistics

Module statistics 

Source

Functions§

chi_square_sf
Survival function of the chi-square distribution, P(X > x), for df degrees of freedom.
finite_range
Inclusive range (min, max) of a sample as f64, or None when empty or entirely non-finite.
g_test_statistic
G-test (likelihood-ratio) statistic for a 2 x k contingency table built from two histograms, together with its degrees of freedom.
hellinger_distance
Hellinger distance, bounded by 1.
histogram_counts
Counts a sample into bins equal-width buckets spanning [min, max].
js_divergence
Jensen-Shannon divergence in nats, bounded by ln 2.
kl_divergence
Kullback-Leibler divergence KL(p || q) in nats. Both inputs must be smoothed probability mass functions of equal length.
kolmogorov_sf
Survival function of the Kolmogorov distribution, Q(z) = 2 * sum_{k>=1} (-1)^(k-1) exp(-2 k^2 z^2).
ks_statistic
Two-sample Kolmogorov-Smirnov statistic D = max_x |F_a(x) - F_b(x)|.
ks_two_sample_p
Asymptotic p-value for a two-sample Kolmogorov-Smirnov statistic d observed on samples of size n1 and n2.
mean
Arithmetic mean, or None for an empty sample.
median
Convenience wrapper: median of a borrowed sample (clones into a scratch buffer so the caller’s data is left untouched).
median_in_place
True median (mean of the two central order statistics for even n), computed with select_nth_unstable_by. Reorders values in place.
normal_two_sided_p
Two-sided normal p-value for a z statistic: P(|Z| > |z|).
population_variance
Population variance (divides by n), or None for an empty sample.
quantile_in_place
Nearest-rank quantile of p in [0, 1], computed in O(n) expected time with select_nth_unstable_by (no full sort). Reorders values in place.
sample_std_dev
Sample standard deviation (divides by n - 1), or None when n < 2.
smoothed_pmf
Normalises counts into a probability mass function with additive (Laplace) smoothing, so that no bucket is ever exactly zero. Zero-probability buckets would make KL divergence infinite and the log-ratio undefined; the smoothing constant is applied symmetrically to both distributions being compared, which is the standard treatment.
sort_ascending
Sorts a slice ascending using the NaN-safe total order.
standard_normal_sf
Survival function of the standard normal distribution, P(Z > z).
total_order
Total ordering for floating-point values that never panics.
wasserstein_1d
First Wasserstein distance (a.k.a. Earth Mover’s Distance) between two one-dimensional empirical distributions.