R/bekk_mc_eval.R

In BEKKs: Multivariate Conditional Volatility Modelling and Forecasting

#' @import future
#' @import future.apply

bekk_mc_eval <- function(object, spec, sample_sizes, iter, nc = 1) {
  xx <- process_object(object)
  theta <- xx$theta

  mse <- rep(NA, length(sample_sizes))
  index <- 1

  for(j in sample_sizes) {
    print(paste('Sample size: ', j))
    sim_dat <- vector(mode = "list", iter)

    sim_dat <- future_lapply(1:iter, function(x){simulate(object, nobs = j)}, future.seed=TRUE)

    dd <- future_lapply(sim_dat, function(x){bekk_fit(spec = spec, data = x, max_iter = 200)},future.seed=TRUE)

    mse[index] <- sum(unlist(lapply(dd, RMSE, theta_true = theta)))/iter
    print(mse[index])
    index <- index +1

  }

  result <- data.frame(Sample_size = sample_sizes, MSE = mse)
  colnames(result) <- c('Sample', 'MSE')
  result <- list(result)

  class(result) <- 'bekkMC'
  return(result)
}


RMSE <- function(x, theta_true) {
  theta_est <- x$theta
  return(mean(sqrt(((theta_true - theta_est) / theta_true)^2)))
}

plot.bekkMC <- function(x, ...) {
  Sample <- NULL
  MSE <- NULL

  msep <- x[[1]]
  ggplot(msep) + geom_line(aes(x = Sample, y = MSE)) + theme_bw()
}