raghat1: Filter a time series with options to shift and scale

In mixAR: Mixture Autoregressive Models

Filter a time series with options to shift and scale

Description

Filter a time series with options to shift and scale. This function is used by mixFilter.

Usage

raghat1(filter, x, index, shift = 0, residual = FALSE, scale = 1)

Arguments

filter	The coefficients of the filter, numeric, see Details.
x	time series, numeric.
index	indices for which to compute the filtered values, numeric.
shift	a constant to be added to each filtered element, a number.
residual	if TRUE calculate a ‘residual’, otherwise calculate a ‘hat’ value.
scale	if scale != 1 calculate scaled residuals by divividing by this value. Probably meaningful only if residual=TRUE.

Details

This function is used by mixFilter. Applies an autoregressive filter to a time series for indices specified by index.

Note that ‘filter’ here is equivalent to calculating one-step predictions (or residuals if residual=TRUE) from autoregressions.

index should not specify indices smaller than length(filter)+1 or larger than length(x)+1. The value length(x)+1 can legitimately be used to calculate a prediction (but not a residual of course) for the first value after the end of the series.

Value

A numeric vector of length equal to length(index).

Note

This should probably use filter but for the purposes of this package filter is usually short and the calculation is vectorised w.r.t. index, so should not be terribly slow.

mixAR documentation built on May 29, 2024, 6:21 a.m.