AFTER1: Forecast combination by AFTER.
In weiqian1/AIafter: Forecast Combination Using the AI-AFTER Algorithm

Description Usage Arguments Details Value Author(s) References See Also Examples

View source: R/AFTER1.R

This function generates forecast combination results using the AFTER algorithm (Yang, 2004).

1	AFTER1(X, y, Xnew, ynew = NULL, n0 = 5)

`X`	Matrix of candidate forecasts for observed series.
`y`	Observed series.
`Xnew`	Matrix of candidate forecasts for unobserved series.
`ynew`	(Optional) True values of unobserved series, usually supplied for testing forecast accurary. If `ynew` is given, a forecast from `Xnew` can use previous observations in `ynew` as in any real forecasting scenarios.
`n0`	Number of “burn-in” observations for AFTER.

This function generates forecasts from the AFTER algorithm. For detailed updating scheme, see Yang (2004).

It returns a list that contains combined forecasts (both fitted values from X and prediction from Xnew):

`fcst`	a forecast matrix, with first column being forecasts from L1-AFTER and second column from L2-AFTER.
`L1Wt`	weight matrix used for L1-AFTER.
`L2Wt`	weight matrix used for L2-AFTER.

Wei Qian, Craig A. Rolling, Gang Cheng and Yuhong Yang
Maintainer: Wei Qian <weiqian@udel.edu>

Yang, Y. (2004), “Combining Forecasting Procedures: Some Theoretical Results,” Econometric Theory, 20, 175-222.

Qian, W., Rolling, C. A., Cheng, G., and Yang, Y. (2021), “Combining Forecasts for Universally Optimal Performance”, International Journal of Forecasting, to appear.

AIafter.

library(AIafter)
data(rgdp)


# observed series and its candidate forecasts
neval <- 20
n <- nrow(rgdp)
y <- rgdp[1:(n-neval),1]
X <- rgdp[1:(n-neval),-1]
# candidate forecasts for unobserved series
Xnew <- rgdp[(n-neval+1):n,-1]

# generate forecasts by AFTER (both fitted and predicted values)
fcst <- AFTER1(X,y,Xnew)

# true values of unobserved series
ynew <- rgdp[(n-neval+1):n,1]
# generate forecasts by AFTER (usually used for evaluation of forecast accuracy)
fcst <- AFTER1(X,y,Xnew,ynew)