Package 'forecastADAPT' reference manual

Title:	Computation of Adaptive Forecast
Description:	The function forAD() implements the adaptive forecasting procedure of Giraitis, Kapetanios and Price (2013) <doi:10.1016/j.jeconom.2013.04.003>. The method can be iterated (e.g., adapt²) and combined with autoregressive (AR) forecasting. These approaches are computationally simple and adapt automatically to structural changes without requiring prior specification of the underlying data-generating process. They are applicable to both stationary and non-stationary time series. The numerical and graphical outputs assist in selecting an appropriate forecasting method, particularly one that minimises mean squared forecast error (MSFE) and yields uncorrelated forecast errors.
Authors:	Violetta Dalla [aut, cre], Liudas Giraitis [aut], George Kapetanios [aut]
Maintainer:	Violetta Dalla <[email protected]>
License:	GPL-3
Version:	0.1.0
Built:	2026-05-08 17:17:06 UTC
Source:	https://github.com/cran/forecastADAPT

Computing adaptive forecast

Description

The functions forAD computes the adaptive one-step ahead forecast of Giraitis, Kapetanios and Price (2013).

Usage

forAD(x, p_max = 3, T0 = 50, n_v = 100, n_AR = 400,
      plots = TRUE, P = 50, PL = 500, p = 1, date_1 = NULL)
forAD(x, p_max = 3, T0 = 50, n_v = 100, n_AR = 400,
      plots = TRUE, P = 50, PL = 500, p = 1, date_1 = NULL)

Arguments

x

A univariate numeric time series (ts, xts, zoo) object or a numeric vector or a numeric data frame column.

p_max

Maximum order of the AR model. Default is 3. If set to 0, no AR is estimated.

T0

Length of the training period used to select tuning parameter $\rho$ . Default is 50.

n_v

Length of the period of forecast errors used in calculation of MSFE. Default is 100.

n_AR

Length of the period of forecast errors used for estimation of AR and for computing correlograms. Default is 400.

plots

Logical. If TRUE (default), the following plots are produced:
1) Rolling MSFEs (adapt, adapt+AR( $p$ ), adapt $^2$ , adapt $^2$ +AR( $p$ ), $p=1,...,p_{max}$ ),
2) Rolling tuning parameters $\rho$ (adapt, adapt $^2$ ),
3) Rolling one-step ahead forecasts (adapt, adapt+AR( $p$ ), adapt $^2$ , adapt $^2$ +AR( $p$ ) for selected $p$ ) and the data,
4) Rolling forecast errors (adapt, adapt+AR( $p$ ), adapt $^2$ , adapt $^2$ +AR( $p$ ) for selected $p$ ),
5) Correlograms of the forecast errors.
Alternatively, plots can be a logical vector specifying which plots, 1)–5), to display.

P

Length of the period used in plots. Default is 50.

PL

Length of the longer period used in plots. Default is 500.

p

Selected AR order used in plots. Default is 1.

date_1

The date for the one-step ahead forecast when x has dates. Default is NULL.

Details

The method implements the adaptive (one-stage adapt or two-stage adapt $^2$ ) one-step ahead forecasting with weights $w_{j,\rho}\propto\rho^j$ for $0<\rho\leq 1$ (normalized to sum up to 1) in the 1st step, combined with AR forecasting in 2nd step.

Value

An object of class 'forAD', which is a list with the following components:

for_1

The one-step ahead forecast for all methods (adapt, adapt+AR( $p$ ), adapt $^2$ , adapt $^2$ +AR( $p$ ), $p=1,...,p_{max}$ ).

rho

The tuning parameter for adaptive methods (adapt, adapt $^2$ ).

ar_coef_se

The estimated coefficients and their standard errors for the adapt+AR( $p$ ), $p=1,...,p_{max}$ .

ar_coef_se_sq

The estimated coefficients and their standard errors for the adapt $^2$ +AR( $p$ ), $p=1,...,p_{max}$ .

MSFE

The MSFE for all methods.

for_in

The rolling one-step ahead forecasts for all methods.

rho_in

The rolling tuning parameters for adaptive methods.

MSFE_in

The rolling MSFE for all methods.

err_in

The rolling one-step ahead forecast errors for all methods.

data

The original data.

Note

Missing values are not allowed.
Permitted sample size: $N\geq25+2p_{max}$ when p_max > 0 and $N\geq21$ when p_max = 0.

Author(s)

Violetta Dalla, Liudas Giraitis and George Kapetanios

References

Giraitis L, Kapetanios G, Price S (2013). "Adaptive Forecasting in the Presence of Recent and Ongoing Structural Change." Journal of Econometrics, 177(2), 153-170. doi:10.1016/j.jeconom.2013.04.003.
Giraitis L, Kapetanios G, Mansur M, Price S (2015). "Forecasting Under Structural Change." In J Beran, Y Feng, H Hebbel (eds.), Empirical Economic and Financial Research: Theory, Methods and Practice, pp. 401–419. Springer International Publishing, Cham. doi:10.1007/978-3-319-03122-4_25.

Examples

# Sample x contains 200 independent standard normal data.
x <- rnorm(200)
print(forAD(x, plots = FALSE))
# Set plots to TRUE to see the plots.
print(forAD(x))
# Sample x contains 200 independent standard normal data.
x <- rnorm(200)
print(forAD(x, plots = FALSE))
# Set plots to TRUE to see the plots.
print(forAD(x))

Package 'forecastADAPT'

Help Index