Diebold-Mariano test for osem forecast comparison

This function performs the Diebold-Mariano test to compare the forecast accuracy of the OSEM model against specified comparison methods (e.g., VAR, BVAR) for insample forecasts. The function allows for flexible specification of insample forecasting methods and handles the preparation of forecast errors for the DM test, including adjustments for overlapping forecasts. The results include DM statistics and p-values for each variable and forecast horizon.

Usage

dm_test(
  model,
  insample_model = NULL,
  insample_methods = c("ets", "auto"),
  insample_main_comparison = "ets",
  insample_sample_share = 0.5,
  comparison_methods = c("VAR", "BVAR"),
  parallel.cores = NULL,
  dm.horizons = 8,
  dm.variance = "bartlett",
  dm.alternative = "two.sided",
  dm.power = 2,
  lags = 4,
  grepl_variables = NULL,
  comparison_data = NULL,
  quiet = FALSE
)

Arguments

model: A model of class 'osem' (to be returned by run_model).
insample_model: A model of class 'osem.forecast.insample' (to be returned by forecast_insample) to be used for the insample forecast comparisons. If provided, other options for insample modelling (sample share, methods, etc.) are ignored.
insample_methods: A character vector of methods to use for the insample forecast comparisons. Default is c("ets", "auto"). Note that the method specified in insample_main_comparison must be included in this vector.
insample_main_comparison: A character string specifying the method to be used as the main comparison for the insample forecasts. Default is "ets". Must be one of the methods specified in insample_methods.
insample_sample_share: A numeric value between 0 and 1 specifying the share of the sample to be used for insample forecasting. Default is 0.5.
comparison_methods: A character vector specifying the methods to compare against the insample main comparison method. Default is c("VAR", "BVAR"). Must be a subset of c("VAR", "BVAR", "RW", "ar", "ets", "auto") which correspond to Vector Autoregressions (VAR), Bayesian VAR (BVAR), Random Walks (RW) and Autoregressive models.
parallel.cores: An integer specifying the number of cores to use for parallel processing when running insample forecasts. If NULL, the function will not use parallel processing. Default is NULL.
dm.horizons: An integer specifying the forecast horizons to be used for the Diebold-Mariano test. Must be a single positive integer. Default is 8.
dm.variance: A character string specifying the variance estimator to be used in the Diebold-Mariano test. Default is "bartlett". Must be one of the variance estimators supported by forecast::dm.test().
dm.alternative: A character string specifying the alternative hypothesis for the Diebold-Mariano test. Default is "two.sided". Must be one of "two.sided", "less", or "greater".
dm.power: A numeric value specifying the power to which forecast errors are raised in the Diebold-Mariano test. Default is 2 (squared errors). Must be a single positive number.
lags: An integer specifying the number of lags to be used in the comparison methods. Default is 4.
grepl_variables: A character vector of variable names to be included in the forecast comparisons. If NULL, all variables will be included. Default is NULL.
comparison_data: A data frame containing pre-prepared forecast errors and related information for the DM test. If provided, the function will use this data instead of running the insample forecasts and preparing the data internally. Default is NULL.
quiet: A logical value indicating whether to suppress messages during the execution of the function. Default is FALSE.

Value

A list containing the results of the Diebold-Mariano test for forecast comparisons, including a tibble with DM statistics and p-values for each variable and forecast horizon, as well as the prepared data used for the comparisons. If the insample_model was not provided, the list will also include the insample forecast model used for the comparisons.