fts.criterion: Information Criteria to Estimate the Number of Factors for...

In ottosven/dffm: Approximate Functional Factor Model

Information Criteria to Estimate the Number of Factors for Functional Factor Models

Description

fts.criterion estimates the optimal number of factors (K) and lags (p) for functional factor models based on information criteria. It applies the methodology proposed by Otto and Salish (2024)

Usage

fts.criterion(fdaobj, K.max = NULL, p.max = NULL)

Arguments

fdaobj	An object of class 'fdaobj', typically the output from fda.preprocess or fts.cumAC.
K.max	An optional integer specifying the maximum number of factors to consider. If NULL (default), K.max will be set to the highest number supported by the given 'fdaobj'.
p.max	An optional integer specifying the maximum number of lags to consider. If NULL (default), p.max will be determined based on Schwert's rule of thumb.

Details

Estimates the number of factors K and lags p for the approximate functional factor model using the information criteria proposed in Otto and Salish (2024).

Value

Returns a list containing:

IC.min	A data.frame with the estimated number of factors (K) and lags (p) based on the Bayesian Information Criterion (BIC), Hannan-Quinn Criterion (HQC), and the final prediction error (fFPE) criterion of Aue et al. (2015).
MSE.matrix	A matrix of Mean Squared Error (MSE) values for all combinations of considered factors and lags.

References

• Otto, S., & Salish, N. (2024). Approximate Factor Models For Functional Time Series. arXiv:2201.02532.

• Aue, A., Norinho, D. D., & Hörmann, S. (2015). On the prediction of stationary functional time series. Journal of the American Statistical Association, 110(509), 378-392.

Examples

fed = load.fed()
fpcaobj = fda.preprocess(data = fed)
criterion_output = fts.criterion(fpcaobj)
print(criterion_output$IC.min)

ottosven/dffm documentation built on Feb. 23, 2025, 1:15 p.m.