rugarch_fit_impl: FIT - GARCH -----

Description

#' Low-Level GARCH function for translating modeltime to forecast #' #' @param formula A dataframe of xreg (exogenous regressors) #' @param data A numeric vector of values to fit #' @param a The order of the non-seasonal auto-regressive (AR) terms. Often denoted "p" in pdq-notation. #' @param g The order of the non-seasonal auto-regressive (AR) terms. Often denoted "p" in pdq-notation. #' @param ... Additional arguments passed to forecast::Arima #' #' @export garch_fit_impl <- function(formula, data, a = 1, g = 1, ar_no_apply = NULL, ma_no_apply = NULL, period = "auto", ...)

# X & Y
others <- list(...)

y <- all.vars(formula)[1] x <- attr(stats::terms(formula, data = data), "term.labels")

outcome <- data[[y]] predictors <- data %>% dplyr::select(dplyr::all_of(x))

# INDEX & PERIOD # Determine Period, Index Col, and Index index_tbl <- modeltime::parse_index_from_data(predictors) period <- modeltime::parse_period_from_index(index_tbl, period) idx_col <- names(index_tbl) idx <- timetk::tk_index(index_tbl)

# XREGS # Clean names, get xreg recipe, process predictors # xreg_recipe <- create_xreg_recipe(predictor, prepare = TRUE) # xreg_matrix <- juice_xreg_recipe(xreg_recipe, format = "matrix")

# FIT outcome <- stats::ts(outcome, frequency = period)

fit_garch <- tseries::garch(outcome, order = c(a, g), ...)

# RETURN modeltime::new_modeltime_bridge( class = "garch_fit_impl",

# Models models = list( model_1 = fit_garch ),

# Data - Date column (matches original), .actual, .fitted, and .residuals columns data = tibble::tibble( !! idx_col := idx, .actual = as.numeric(outcome), .fitted = fit_garch$fitted.values[,1], .residuals = fit_garch$residuals ),

extras = list( y_var = y, period = period, otros = others ),

# Description - Convert arima model parameters to short description desc = stringr::str_glue('GARCH ({fit_garch$order[1]}, {fit_garch$order[2]}) Model') )

Usage

rugarch_fit_impl(
  formula,
  data,
  a = 1,
  g = 1,
  ar = 1,
  ma = 1,
  tune_by = NULL,
  period = "auto",
  ...
)

Arguments

formula

A dataframe of xreg (exogenous regressors)

data

A numeric vector of values to fit

The order of ARCH part

The order of GARCH part

The order of the non-seasonal auto-regressive (AR) terms. Often denoted "p" in pdq-notation.

tune_by

Parameter for tuning.

period

Period

...

Additional arguments passed to forecast::Arima

Value

A fitted model

Details

#' @export print.garch_fit_impl <- function(x, ...) print(x$models$model_1) invisible(x)

Low-Level GARCH function for translating modeltime to forecast