predict_rkhs function

Predict functional time series using ARH RKHS.

using an ARH of order 1 obtain 1 step ahead forecast and 1-$alpha$

predictive confidence bands for the forecasted function.

predict_rkhs(
  model,
  newdata,
  bands = FALSE,
  B = 100,
  level = 0.95,
  kvec = round(sqrt(2 * B))
)

Arguments

• model: a arh_rkhs object containing the functional objects and the lambda coefficients of the d dimensional RKHS representation and the autocorrelation operator.
• newdata: an optional data frame in which to look for variables with which to predict. If missing, the fitted values are used.
• bands: logical variable indicating if the predictive confidence band is computed. Default = FALSE.
• B: number of bootstrap replicates for the band construction. Needed if bands = TRUE. Default = 100.
• level: confidence level for the band construction. Needed if bands = TRUE. Default = 0.95.
• kvec: number of neighbour points to consider in the computation of the minimum entropy set.

Returns

• forecast: 1 step ahead forecast. - fitted: fitted values. - UB: upper bound of the 1-$alpha$ predictive confidence band. - LB: lower bound of the 1-$alpha$ predictive confidence band. - bootsrap.pred: bootstrap pseudo replicates.

• bootsrap.pred.inband: bootstrap pseudo replicates included in the 1-$alpha$ predictive confidence band. - res: estimation residuals.

References

N. Hernández, J. Cugliari, J. Jacques. Simultaneous Predictive Bands for Functional Time Series using Minimum Entropy Sets. arXiv:2105.13627 (2021).

Author(s)

N. Hernández and J. Cugliari