Home › Stats-TimeSeries › Serial correlation

LjungBoxResult

A Ljung-Box test at each lag, indexed from lag 1.

public sealed class LjungBoxResult

Properties — Statistics is the Ljung-Box statistic cumulated to each lag. PValues is its chi-square p-value, NaN where no degree of freedom is left. BoxPierceStatistics is the Box-Pierce statistic, empty unless LjungBoxOptions.BoxPierce asked for it. BoxPiercePValues is Box-Pierce's p-value, empty unless it was asked for. DegreesOfFreedom is the lag less the model's parameters, which may be zero or negative.

Example — five parallel lists rather than a list of five-field rows, because a caller plots a column.

using Lodestar.Stats.TimeSeries;

double[] series = [1.0, 3.0, 2.0, 5.0, 4.0, 7.0, 6.0, 9.0, 8.0, 11.0, 10.0, 13.0];

LjungBoxResult result = SerialCorrelation.LjungBox(
    series, lagCount: 4, new LjungBoxOptions { BoxPierce = true });

int lastLagDf = result.DegreesOfFreedom[3];                       // => 4
double lastStatistic = Math.Round(result.Statistics[3], 4);       // => 11.7928
double lastBoxPierce = Math.Round(result.BoxPierceStatistics[3], 4);  // => 8.6989

Remarks — there is no public constructor. A result is what SerialCorrelation.LjungBox returns, never something a caller assembles by hand.

BoxPierceStatistics and BoxPiercePValues are empty lists, not lists of NaN, when LjungBoxOptions.BoxPierce was left at its false default — a reader indexing them by mistake gets an ArgumentOutOfRangeException rather than a silent NaN.

A class rather than a record, for the reason AutocorrelationResult gives: a record's equality would compare these five lists by reference.

Applies to — net10.0, netstandard2.0.

See also — SerialCorrelation.LjungBox, LjungBoxOptions, the Python equivalence table.