predict.tsgarch.estimate {tsgarch} | R Documentation |
Model Prediction
Description
Prediction function for class “tsgarch.estimate”.
Usage
## S3 method for class 'tsgarch.estimate'
predict(
object,
h = 1,
newxreg = NULL,
newvreg = NULL,
nsim = 0,
sim_method = c("parametric", "bootstrap"),
block = 1,
forc_dates = NULL,
init_states = NULL,
seed = NULL,
...
)
Arguments
object |
an object of class “tsgarch.estimate”. |
h |
the forecast horizon. |
newxreg |
not currently used, |
newvreg |
variance regressors rows equal to h. This can be either a numeric or xts matrix. Only needed if the model was estimated with regressors in the variance equation. |
nsim |
the number of simulations to use for generating the simulated predictive distribution. Defaults to zero (no simulated distribution). |
sim_method |
the simulation method to use when nsim great than zero. The “parametric” method samples from the model distribution whilst the “bootstrap” from the standardized model residuals. |
block |
for the “bootstrap” sim_method, this allows to generate block length samples (defaults to 1). |
forc_dates |
an optional vector of forecast dates equal to h. If NULL will use the implied periodicity of the data to generate a regular sequence of dates after the last available date in the data. |
init_states |
an optional vector of states to initialize the forecast. If NULL, will use the last available state from the estimated model. This must be equal to the max of the ARCH and GARCH terms. |
seed |
an integer that will be used in a call to set.seed before simulating. |
... |
additional arguments for future expansion options. |
Details
The bootstrap method considered here, is based on re-sampling innovations from the empirical distribution of the fitted GARCH model to generate future realizations of the series and sigma. This only considers distributional uncertainty and will not generate prediction intervals for the 1-step ahead sigma forecast for which only the parameter uncertainty is relevant in GARCH type models (and not currently implemented). When the horizon h is equal to 1, no simulation is performaed since there is no uncertainty to account for.
Value
A “tsgarch.predict” object.
References
Pascual,L., Romo,J., Ruiz,E. (2006). “Bootstrap prediction for returns and volatilities in GARCH models.” Computational Statistics & Data Analysis, 50(9), 2293–2312.