op.arima {popstudy}

op.arima

Description

Estimates the best predictive ARIMA model using overparameterization.

Usage

op.arima(
  arima_process = c(p = 1, d = 1, q = 1, P = 1, D = 1, Q = 1),
  seasonal_periodicity,
  time_serie,
  reg = NULL,
  horiz = 12,
  prop = 0.8,
  training_weight = 0.2,
  testing_weight = 0.8,
  parallelize = FALSE,
  clusters = detectCores(logical = FALSE),
  LAMBDA = NULL,
  ISP = 100,
  ...
)

Arguments

arima_process

numeric. The ARIMA(p,d,q)(P,D,Q) process.

seasonal_periodicity

numeric. The seasonal periodicity, 12 for monthly data.

time_serie

ts. The univariate time series object to estimate the models.

reg

Optionally, a vector or matrix of external regressors, which must have the same number of rows as time_serie.

horiz

numeric. The forecast horizon.

prop

numeric. Data proportion for training dataset.

training_weight

numeric. Importance weight for the goodness of fit and precision measures in the training dataset.

testing_weight

numeric. Importance weight for the goodness of fit and precision measures in the testing dataset.

parallelize

logical. If TRUE, then use parallel processing.

clusters

numeric. The number of clusters for the parallel process.

LAMBDA

Optionally. See forecast::Arima() for details.

ISP

numeric. Overparameterization indicator to filter the estimated models in the (0,100] interval.

...

additional arguments to be passed to forecast::Arima().

Value

op.arima returns an object of class list with the following components:

arima_models

all models defined by the arima_process argument.

final_measures

goodness of fit and precision measures for each model.

bests

a sorted list with the best ARIMA models.

best_model

a list of "Arima", see forecast::Arima()

Author(s)

Cesar Gamboa-Sanabria

References

Gamboa-Sanabria C (2022). La Sobreparametrizacion en el ARIMA: una aplicacion a datos costarricenses. Master's thesis, Universidad de Costa Rica.

Examples

op.arima(arima_process = c(2,1,2,2,1,2),
time_serie = AirPassengers,
seasonal_periodicity = 12, parallelize=FALSE)