ARFIMAforecast-class: class: ARFIMA Forecast Class

ARFIMAforecast-classR Documentation

class: ARFIMA Forecast Class

Description

Class for the ARFIMA forecast.

Slots

forecast:

Object of class "vector"

model:

Object of class "vector"

Extends

Class "ARFIMA", directly. Class "rGARCH", by class "ARFIMA", distance 2.

Methods

fitted

signature(x = "ARFIMAforecast"): The n.ahead by n.roll+1 matrix of conditional mean forecasts. The column names are the T[0] dates.

fpm

signature(object = "ARFIMAforecast"): Forecast performance measures.

show

signature(object = "ARFIMAforecast"): Forecast summary returning the 0-roll frame only.

Note

Since versions 1.01-3, the fitted method has been introduced which extracts the n.ahead by (n.roll+1) matrix of conditional mean forecasts, with column names the T[0] time index. This is unlike the old data.frame which returned the T+1 etc dates. This method is the default extractor in rugarch package for the conditional mean (whether from an estimated, filtered, forecast or simulated object) and the other method, namely as.data.frame is now deprecated with the exception of a few classes where it is still used (ARFIMAdistribution and ARFIMAroll).
The fpm method returns the Mean Squared Error (MSE), Mean Absolute Error (MAE), Directional Accuracy (DAC) and number of points used for the calculation (N), of forecast versus realized returns, if the extra summary option is set to TRUE (default). This is a 4 x (n.roll+1) matrix, with row headings the T[0] time index, and requires at least 5 points to calculate the summary measures else will return NA. When n.ahead>1, this method calculates the measures on the n.ahead>1 unconditional forecast, but if n.ahead=1 with n.roll>4, it will calculate the measures on the rolling forecast instead. Finally, when summary is set to FALSE, the method will return a list of length n.roll+1 of xts objects with the loss functions (Squared Error and Absolute Error and Directional Hits).

Author(s)

Alexios Ghalanos


rugarch documentation built on Sept. 20, 2023, 9:07 a.m.