Stats andersonksamplevariant

Development build. This page describes main, not a released package. The latest published Lodestar.Stats is 0.5.0 — read its documentation.

Home › Stats › Hypothesis tests

AndersonKSampleVariant

Which form of the k-sample Anderson-Darling statistic AndersonDarling.KSample computes: scipy 1.18's variant.

public enum AndersonKSampleVariant { Midrank, Right, Continuous }

Members — Midrank, the default, is Scholz and Stephens' equation 7, with mid-ranks for ties, for continuous and discrete samples alike: scipy's midrank=True. Right is their equation 6, for discrete samples: midrank=False. Continuous is equation 3, which assumes no ties.

Example — the same two samples under two variants.

using Lodestar.Stats;

double[] before = [2.1, 3.4, 1.9, 4.2, 3.3, 2.8, 3.9, 2.2, 3.1, 2.7];
double[] after = [2.5, 3.1, 2.8, 4.6, 3.9, 3.0, 4.4, 2.9, 3.5, 3.6];

double midrank = Math.Round(AndersonDarling.KSample(before, after).Statistic, 6);
double continuous = Math.Round(
    AndersonDarling.KSample(AndersonKSampleVariant.Continuous, before, after).Statistic, 6);

double withTies = midrank;          // => 0.385197
double withoutTies = continuous;    // => 0.342788

Remarks — scipy 1.18 is replacing midrank with variant, and will drop the critical values from the result it returns; these are kept, since they are what a clamped p-value leaves to read.

Applies to — net10.0, netstandard2.0.

See also — AndersonDarling.KSample.