Bootstrapped Confidence Intervals — confidence

Method for the computation of bootstrapped confidence intervals.

Usage

confidence_intervals(
  data,
  pred_model,
  h,
  train = 0.8,
  M = 1000,
  alpha = 0.05,
  output = NULL,
  ...
)

Arguments

data: Data.frame (or vector) to apply cross-validation where the rows are the observations and the columns are the variables.
pred_model: Model to predict the data where the first argument in the data and the second the length of prediction. Additional parameters can be passed using ... . Returns a vector of the predictions.
h: Number of observations to predict into the future.
train: Minimum amount of data (0,1) to train model. After this point, a sliding window approach is taken.
M: Numeric. Number of bootstrap iterations.
alpha: Significance level for the intervals
output: Numeric or name of the column for the output when using a data.frame for data. If not given, the first column is used.
...: Additional parameters to pass to pred_model.

Value

Data.frame with columns for the forecasts and lower/upper confidence intervals.

Examples

data <- cumsum(rnorm(150))
pred_model <- function(x, h) {
  predict(forecast::ets(x), h = h)$mean
}
h <- 10
ets_model <- predict(forecast::ets(data[1:140]), h)
#> Registered S3 method overwritten by 'quantmod':
#>   method            from
#>   as.zoo.data.frame zoo 
ints <- confidence_intervals(data = data[1:140], pred_model = pred_model, h = h)

plot(ets_model)
lines(x = 141:150, y = ints$lower, col = "red")
lines(x = 141:150, y = ints$upper, col = "red")