Lodestar.Stats 0.4.0. This page is frozen at that release. Read the current documentation for what main says now. A link to a decision or a migration page follows main, and leaves the archive.

Lodestar.Stats

Ten families of classical hypothesis test, at scipy.stats 1.18.0 parity. Arrays in, a statistic and a p-value out; nothing is fitted, so every entry point is static.

test	what it asks	entry point
Student / Welch t	do two samples have the same mean?	TTest
Mann-Whitney U	the same question, assuming no shape	MannWhitney
Wilcoxon signed-rank	the same, on paired measurements	Wilcoxon
χ²	do counts match an expected distribution, or are two factors independent?	ChiSquare
Fisher exact	the same for a 2×2 table, at any sample size	FisherExact
Kolmogorov-Smirnov	do two samples share a distribution?	KolmogorovSmirnov
one-way ANOVA	do several groups share one mean?	OneWayAnova
Kruskal-Wallis	the same, assuming no shape	KruskalWallis
Shapiro-Wilk	could this sample be normal?	ShapiroWilk
Bonferroni / BH / BY	how many of these results are chance?	MultipleComparisons

Every family but three returns the same two numbers, TestResult; a t-test also carries its degrees of freedom (TTestResult), a contingency table also carries the table independence would have produced (Chi2ContingencyResult), and Kolmogorov-Smirnov also carries where and in which direction the two samples parted furthest (KsResult).

result	carries	returned by
TestResult	a statistic, a p-value	eight of the ten families
TTestResult	+ degrees of freedom, a confidence interval	TTest
Chi2ContingencyResult	+ degrees of freedom, the expected table	ChiSquare.Contingency
KsResult	+ where the gap was reached, and its sign	KolmogorovSmirnov

Five small enums choose what a test asks, and how it answers when the exact and the approximate route disagree.

option	chooses
Alternative	which tail the p-value covers
Variance	whether an independent-samples t-test pools the two variances
Continuity	whether a discrete statistic's normal approximation gets the half-unit correction
ExactMethod	the exact null distribution, its normal approximation, or a choice between them
ZeroMethod	what Wilcoxon does with a pair whose difference is exactly zero

The hypothesis-testing guide says which test answers which question, and the Python equivalence table maps each scipy call to its counterpart here.