Development build. This page describes main, not a released package. The latest published Lodestar.Stats is 0.4.0 — read its documentation.

Home › Stats › Overview

NanPolicy

What a test does with a NaN in its input; scipy's nan_policy.

public enum NanPolicy

Fields — Propagate carries the NaN into the statistic and the p-value, which is scipy's default and this package's behaviour before the parameter existed. Raise refuses the input with ArgumentException. Omit drops the missing observations and tests on the rest.

Example — the same sample under two policies.

using Lodestar.Stats;

double[] sample = [1.0, 2.0, double.NaN, 4.0, 5.0, 6.0, 7.0];

TestResult carried = ShapiroWilk.Test(sample);
bool isNaN = double.IsNaN(carried.Statistic);  // => True

TestResult dropped = ShapiroWilk.Test(sample, NanPolicy.Omit);
bool isReal = double.IsNaN(dropped.Statistic);  // => False

Remarks — offered on the eleven entry points whose scipy counterpart takes nan_policy, and on no others: chi2_contingency, fisher_exact and false_discovery_control do not take it, so ChiSquare.Contingency, FisherExact.Test and the MultipleComparisons methods do not either.

Omit drops pairs where the two inputs are aligned — TTest.Paired, Wilcoxon.Paired, and ChiSquare.GoodnessOfFit when an expectation is given — and values everywhere else. Dropping each sample independently would silently change what a paired test is testing.

Omission is a filter, not a second policy: a family's own guards run afterwards on what survives, so ShapiroWilk.Test still refuses fewer than three observations and KruskalWallis.Test still refuses a fully tied pool. Decision 0007 has the reasoning.

Applies to — net10.0, netstandard2.0.

See also — ShapiroWilk.Test, TTest.Paired.