forecast_roll: Out-of-sample Forecasting based on Rolling Window

Description

This function conducts rolling window forecasting.

Usage

forecast_roll(object, n_ahead, y_test, num_thread = 1, ...)
# S3 method for bvharcv
print(x, digits = max(3L, getOption("digits") - 3L), ...)
is.bvharcv(x)
# S3 method for bvharcv
knit_print(x, ...)
# S3 method for olsmod
forecast_roll(object, n_ahead, y_test, num_thread = 1, ...)
# S3 method for normaliw
forecast_roll(object, n_ahead, y_test, num_thread = 1, use_fit = TRUE, ...)
# S3 method for ldltmod
forecast_roll(
  object,
  n_ahead,
  y_test,
  num_thread = 1,
  level = 0.05,
  stable = FALSE,
  sparse = FALSE,
  med = FALSE,
  lpl = FALSE,
  use_fit = TRUE,
  verbose = FALSE,
  ...
)
# S3 method for svmod
forecast_roll(
  object,
  n_ahead,
  y_test,
  num_thread = 1,
  level = 0.05,
  use_sv = TRUE,
  stable = FALSE,
  sparse = FALSE,
  med = FALSE,
  lpl = FALSE,
  use_fit = TRUE,
  verbose = FALSE,
  ...
)

Value

predbvhar_roll

class

Arguments

object: Model object
n_ahead: Step to forecast in rolling window scheme
y_test: Test data to be compared. Use divide_ts() if you don't have separate evaluation dataset.
num_thread: Number of threads
...: not used
x: Any object
digits: digit option to print
use_fit: Use object result for the first window. By default, TRUE.
level: Specify alpha of confidence interval level 100(1 - alpha) percentage. By default, .05.
stable: Filter only stable coefficient draws in MCMC records.
sparse: Apply restriction. By default, FALSE.
med: If TRUE, use median of forecast draws instead of mean (default).
lpl: Compute log-predictive likelihood (LPL). By default, FALSE.
verbose: Print the progress bar in the console. By default, FALSE.
use_sv: Use SV term

Details

Rolling windows forecasting fixes window size. It moves the window ahead and forecast h-ahead in y_test set.

References

Hyndman, R. J., & Athanasopoulos, G. (2021). Forecasting: Principles and practice (3rd ed.). OTEXTS.