Home › Stats-TimeSeries › Seasonality

SeasonalDecompositionOptions

What a seasonal decomposition may be told.

public sealed record SeasonalDecompositionOptions

Properties — Model is additive or multiplicative, a SeasonalModel; Additive by default. TwoSided says whether the moving average is centred, true by default, or trails the point. ExtrapolateTrend is how many of the nearest defined trend points, less one, fit the lines that fill the trend's undefined ends; 0 by default, which leaves them NaN.

Exceptions — ArgumentOutOfRangeException when ExtrapolateTrend is negative, or Model is a value SeasonalModel does not declare, checked where the value is set.

Example — the ends filled by lines through the nearest two defined points.

using Lodestar.Stats.TimeSeries;

double[] quarterly = [10.0, 14.0, 8.0, 12.0, 11.0, 15.0, 9.0, 13.0, 12.0, 16.0, 10.0, 14.0];

SeasonalComponents filled = SeasonalDecomposition.Decompose(
    quarterly, 4, new SeasonalDecompositionOptions { ExtrapolateTrend = 1 });

double firstTrend = Math.Round(filled.Trend[0], 4);  // => 10.625
double lastTrend = Math.Round(filled.Trend[11], 4);  // => 13.375

Remarks — the reference's extrapolate_trend="period" is period − 1 here. Its back window leaves the last defined point out of the fit, and so does this one: the corpus freezes that.

Being a record of value types, two option sets with the same three values are equal.

Applies to — net10.0, netstandard2.0.

See also — SeasonalDecomposition.Decompose, SeasonalComponents.