choose.threshold: Search the optimal threshold ratio used for model averaging
In xqnwang/fuma: Forecast uncertainty based on model averaging (FUMA)

Description Usage Arguments Details Value

According to the fitted values of the pre-trained algorithm, the optimal threshold ratio used for model averaging is searched from a set of pre-set threshold ratios.

choose.threshold(
  dataset,
  fitProb,
  ratio,
  combine = "mean",
  level,
  parallel = FALSE,
  num.cores = 2
)

`dataset`	a list of time series. See details for the required format.
`fitProb`	a matrix with `n` (number of time series) rows and `F` (number of benchmark methods) matrix. The matrix is applied to assign weights to each benchmark method for each time series.
`ratio`	a vector with pre-set threshold ratios. The threshold ratio quantifies as a number between 0 and 1, where 0 indicates all the benchmark methods are selected and 1 indicates only the benchmark method with the minimal value of scores is selected.
`combine`	string representing the interval combination methods. "mean", "weightedmean" or "median".
`level`	the confidence level of prediction intervals that are used for combination.
`parallel`	logical. If `TRUE` then the calculations are conducted in parallel.
`num.cores`	the specified amount of parallel processes to be used if parallel = TRUE.

dataset must be a list with each element having the following format:

n: the number of observations in the time series.
h: the number of required forecasts.
period: interval of the time series. Possible values are "YEARLY", "QUARTERLY", "MONTHLY" & "OTHER".
x: a time series of length n (the historical data).
xx: a time series of length h (the future data).
ff: a matrix with F rows and h columns. Each row contains the forecasts of each method in methods.
lower: a list with each element being the matrix of lower bounds for certain confidence level.
upper: a list with each element being the matrix of upper bounds for certain confidence level.

A list with the following elements

threshold_y: numeric, the optimal threshold ratio searched for yearly data.
threshold_q: numeric, the optimal threshold ratio searched for quarterly data.
threshold_m: numeric, the optimal threshold ratio searched for monthly data.
msis_y: a matrix of MSIS values for each pre-set threshold ratio to yearly data.
msis_q: a matrix of MSIS values for each pre-set threshold ratio to quarterly data.
msis_m: a matrix of MSIS values for each pre-set threshold ratio to monthly data.