ets_modelspec: Model Specification

In tsmodels/tsets: Exponential Smoothing Models

Model Specification

Description

Specifies an ETS model prior to estimation.

Usage

ets_modelspec(
  y,
  model = "AAN",
  damped = FALSE,
  power = FALSE,
  xreg = NULL,
  frequency = NULL,
  transformation = "box-cox",
  lambda = NULL,
  normalized_seasonality = TRUE,
  fixed_pars = NULL,
  scale = FALSE,
  seasonal_init = "fixed",
  lower = 0,
  upper = 1,
  sampling = NULL,
  xreg_init = TRUE,
  ...
)

Arguments

y	an xts vector.
model	the type of model (based on the taxonomy in Hyndman) where the first letter denotes the type of error (E), the second letter the type of trend (T) and the third letter the type of seasonality (S). A value of N denotes none. Models supported are “AAA”, “AAN”, “ANN”, “ANA”, “MMM”, “MMN”, “MNN”, “MNM”, “MAM” and “MAN”. In addition, one may set model equal to “Theta”, which is equivalent to a “AAN” model with slope parameter (beta) fixed to zero.
damped	whether to include a damped trend.
power	whether to use the power model (only for multiplicative error with additive trend combination), i.e. the “MAM” and “MAN” models.
xreg	an optional xts matrix of regressors (pre-lagged).
frequency	the frequency of y (if using a seasonal model).
transformation	a valid transformation for y from the “tstransform” function in the “tsaux” package (currently box-cox or logit are available) applied to additive models only.
lambda	the Box Cox power parameter (lambda). If NA will estimate this using the method of Guerrero.
normalized_seasonality	whether to impose Roberts-McKenzie normalized seasonality.
fixed_pars	a named vector of valid parameter names with values which will be fixed rather than estimated. Valid values are as follows: alpha the adjustment coefficient on the Level component beta the adjustment coefficient on the Slope component gamma the adjustment coefficient on the Seasonal component\ phi the damping parameter theta the power exponent for the Level and Slope components in the power model delta the power exponent for the Seasonal component in the power model l0 the initial state value for the Level component b0 the initial state value for the Slope component s0,..., s[m-1] e.g. s11, the initial state values for the Seasonal component rho1,...,rho[k] e.g. rho12, the coefficients on the regressors
scale	whether to rescale the data using y/max(y) (only for additive models). This sometimes helps in the optimization.
seasonal_init	whether the initial seasonal states are estimated or fixed (set to a backcast approximation).
lower	the lower bound for the transformation.
upper	the upper bound for the transformation.
sampling	sampling frequency of the dataset. If NULL, will try to identify from the timestamps of y. This is useful for plotting and extending the timestamps in the prediction horizon.
xreg_init	whether to find initial estimates for the regressors with tighter lower and upper bounds. This is only applicable for additive error models.
...	not used.

Details

The specification object holds the information and data which is then passed to the maximum likelihood estimation routines.

Value

An object of class “tsets.spec”.

tsmodels/tsets documentation built on Oct. 8, 2022, 9:15 a.m.