SBAGM (version 0.1.0)

appgarch: Find the appropriate ARMA-GARCH model

Description

The appgarch function computes RMSE and MAE of the all possible combinations of GARCH type model and distribution, and forecast value. Based on the lowest RMSE and MAE, we can find the best model and distribution combinations of the particular data.

Usage

appgarch(data, methods = c("sGARCH", "gjrGARCH"),
distributions = c("norm", "std", "snorm"), aorder = c(1, 0),
gorder = c(1, 1), algo = "gosolnp", stepahead = 5)

Arguments

data

Univariate time series data

methods

Volatility models. Valid models are <U+201C>sGARCH<U+201D>, <U+201C>eGARCH<U+201D>, <U+201C>gjrGARCH and <U+201C>csGARCH<U+201D>. Default: methods= c("sGARCH", "gjrGARCH").

distributions

The conditional density to use for the innovations. Valid choices are <U+201C>norm<U+201D> for the normal distibution, <U+201C>snorm<U+201D> for the skew-normal distribution, <U+201C>std<U+201D> for the student-t, <U+201C>sstd<U+201D> for the skew-student, <U+201C>ged<U+201D> for the generalized error distribution, <U+201C>sged<U+201D> for the skew-generalized error distribution, <U+201C>nig<U+201D> for the normal inverse gaussian distribution, <U+201C>ghyp<U+201D> for the Generalized Hyperbolic, and <U+201C>jsu<U+201D> for Johnson's SU distribution. Default: distributions= c("norm", "std", "snorm").

aorder

ARMA order. Default: aorder=c(1, 0)

gorder

GARCH order. Default: gorder=c(1, 1)

algo

Solver. One of either <U+201C>nlminb<U+201D>, <U+201C>solnp<U+201D>, <U+201C>lbfgs<U+201D>, <U+201C>gosolnp<U+201D>, <U+201C>nloptr<U+201D> or <U+201C>hybrid<U+201D>. Default: algo = "gosolnp". (see documentation in the rugarch-package for details)

stepahead

The forecast horizon.

Value

rmse_mean

Root Mean Square Error (RMSE) value of the mean forecast for all combinations

mae_mean

Mean Absolute Error (MAE) value of the mean forecast for all combinations

forecast_mean

Mean forecast for all combinations

forecast_sigma

Sigma value for all combinations

Details

It allows for a wide choice in univariate GARCH models, distributions, and mean equation modelling. If the user provides the model combinations like methods= c("sGARCH", <U+201C>eGARCH", gjrGARCH") and distributions combination like distributions= c("norm", "std", "snorm") along with the other parameters, then get the RMSE and MAE value for all possible combinations of methods and distributions, which helps to find the best GARCH type model based on the lowest RMSE and MAE value.

References

Bollerslev, T. (1986). Generalized autoregressive conditional heteroscedasticity. Journal of Econometrics, 31, 307-327.

Engle, R. (1982). Autoregressive conditional heteroscedasticity with estimates of the variance of United Kingdom inflation, Econometrica, 50, 987-1008.

See Also

appmsgarch, ARIMAAIC

Examples

# NOT RUN {
data("ReturnSeries")
appgarch(ReturnSeries)
# }