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.

Distributions.FisherSf

The upper tail of the F distribution: P(F > f).

public static double FisherSf(double f, double numeratorDf, double denominatorDf)

Parameters — f is the statistic. numeratorDf and denominatorDf are the two degrees of freedom, both of which must be positive.

Returns — scipy.stats.f.sf(f, dfn, dfd).

Exceptions — ArgumentOutOfRangeException when either degrees-of-freedom argument is not positive, NaN included.

Example — the overall significance of a model with two regressors on twenty residual degrees of freedom.

using Lodestar.Stats;

double overall = Distributions.FisherSf(4.0, 2.0, 20.0);  // => 0.0345716…

// Far into the tail, where an absolute tolerance would accept a zero.
double extreme = Distributions.FisherSf(500.0, 3.0, 100.0);  // => 4.8466962308084166E-60

Remarks — the same tail one-way ANOVA already reports here, exposed for a caller that computed its own F — a regression's overall test, or a nested-model comparison — rather than handing this package its groups.

The two degrees of freedom are not interchangeable: FisherSf(f, a, b) and FisherSf(f, b, a) are different numbers, and the numerator's is the one that counts the constraints being tested.

Applies to — net10.0, netstandard2.0.

See also — OneWayAnova, Distributions.StudentSf.