Home › Stats-TimeSeries › Stationarity tests

KpssLagRule

How KPSS chooses the lag window of its long-run variance.

public enum KpssLagRule

Fields — Automatic is Hobijn, Franses and Ooms' (1998) window, estimated from the residuals — the reference's "auto" and the default. Legacy is Schwert's ceil(12·(n/100)^¼), the reference's "legacy". Fixed takes KpssOptions.LagCount as given.

Example — three rules, three windows, on one series.

using Lodestar.Stats.TimeSeries;

double[] walk = [0.0, 1.2, 0.7, 2.1, 3.0, 2.4, 3.9, 5.1, 4.6, 6.0, 7.3, 6.8,
                 8.2, 9.5, 9.1, 10.4, 11.8, 11.2, 12.7, 14.0, 13.5, 14.9, 16.2, 15.8];

int automatic = Stationarity.Kpss(walk).LagCount;                                                      // => 3
int legacy = Stationarity.Kpss(walk, new KpssOptions { LagRule = KpssLagRule.Legacy }).LagCount;         // => 9
int fixedWindow = Stationarity.Kpss(walk, new KpssOptions { LagRule = KpssLagRule.Fixed, LagCount = 2 }).LagCount;  // => 2

Remarks — both computed windows are capped at one below the series length.

Applies to — net10.0, netstandard2.0.

See also — KpssOptions, Stationarity.Kpss.