ETSforecast function

ETSforecast

Estimates and forecasts ETS general univariate models

ETSforecast(
  y,
  u = NULL,
  model = "???",
  s = frequency(y),
  h = max(2 * s, 6),
  criterion = "aicc",
  lambda = 1,
  armaIdent = FALSE,
  identAll = FALSE,
  forIntervals = FALSE,
  bootstrap = FALSE,
  nSimul = 5000,
  verbose = FALSE,
  alphaL = c(1e-08, 1 - 1e-08),
  betaL = alphaL,
  gammaL = alphaL,
  phiL = c(0.8, 0.98),
  p0 = -99999
)

Arguments

• y: a time series to forecast (it may be either a numerical vector or a time series object). This is the only input required. If a vector, the additional input s should be supplied compulsorily (see below).

• u: a matrix of input time series. If the output wanted to be forecast, matrix u should contain future values for inputs.

• model: the model to estimate. It is a single string indicating the type of model for each component with one or two letters:

  • Error: ? / A / M
  • Trend: ? / N / A / Ad / M / Md
  • Seasonal: ? / N / A / M

• s: seasonal period of time series (1 for annual, 4 for quarterly, ...)

• h: forecast horizon. If the model includes inputs h is not used, the lenght of u is used instead.

• criterion: information criterion for identification ("aic", "bic" or "aicc").

• lambda: Box-Cox lambda parameter (NULL: estimate)

• armaIdent: check for arma models for error component (TRUE / FALSE).

• identAll: run all models to identify the best one (TRUE / FALSE)

• forIntervals: estimate forecasting intervals (TRUE / FALSE)

• bootstrap: use bootstrap simulation for predictive distributions

• nSimul: number of simulation runs for bootstrap simulation of predictive distributions

• verbose: intermediate estimation output (TRUE / FALSE)

• alphaL: constraints limits for alpha parameter

• betaL: constraints limits for beta parameter

• gammaL: constraints limits for gamma parameter

• phiL: constraints limits for phi parameter

• p0: initial values for parameter search (alpha, beta, phi, gamma) with consraints:

  • 0 < alpha < 1
  • 0 < beta < alpha
  • 0 < phi < 1
  • 0 < gamma < 1 - alpha

Returns

An object of class ETS. It is a list with fields including all the inputs and the fields listed below as outputs. All the functions in this package fill in part of the fields of any ETS object as specified in what follows (function ETS fills in all of them at once):

After running ETSforecast or ETSestim: - p: Estimated parameters

• criteria: Values for estimation criteria (LogLik, AIC, BIC, AICc)

• yFor: Forecasted values of output

• yForV: Variance of forecasted values of output

• ySimul: Bootstrap simulations for forecasting distribution evaluation

After running ETSvalidate: - table: Estimation and validation table

• comp: Estimated components in matrix form

After running ETScomponents: - comp: Estimated components in matrix form

Details

ETSforecast is a function for modelling and forecasting univariate time series with ExponenTial Smoothing (ETS) time series models. It sets up the model with a number of control variables that govern the way the rest of functions in the package will work. It also estimates the model parameters by Maximum Likelihood and forecasts the data.

Examples

## Not run:

y <- log(AirPAssengers)
m1 <- ETSforecast(y)
m1 <- ETSforecast(y, model = "A?A")
## End(Not run)

See Also

ETS, ETSvalidate, ETScomponents, ETSestim

Author(s)

Diego J. Pedregal