dynupdateweightselect {ftsa}R Documentation

Selecting an optimal tuning parameter for the PLS and RR methods

Description

Selecting an optimal tuning parameter for the penalized least squares and ridge regression methods used in the dynupdate.

Usage

dynupdateweightselect(data, method = c("pls", "ridge"), interval = c(0, 10^4), p, 
                      backh = 1, errortype = c("mse", "mae", "mape"))

Arguments

data An object of class sfts.
method Either the penalized least squares method or the ridge regression method.
interval Interval where the optimization takes place.
p The most recent time period within the function support range.
backh The number of curves used as the validation data. For instance, when backh = 2, the last two curves are subsequently treated as the validation data, while the remaining curves are treated as the training data.
errortype Type of error measure.

Details

The optimal tuning parameter is selected by minimizing an averaged forecast error measure. This function also utilizes the optimize function.

Value

minimum Optimal tuning parameter.
objective Value of the error function at the optimal point.

Note

This function can be computationally intensive for backh > 1.

Author(s)

Han Lin Shang

References

H. Shang and R. J. Hyndman (2009) "Nonparametric time series forecasting with dynamic updating", In R. S. Anderssen, R. D. Braddock and L.T.H. Newham (eds), 18th World IMACS Congress and MODSIM09 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand and International Association for Mathematics and Computers in Simulation, July 2009, pp. 1552-1558. ISBN: 978-0-9758400-7-8. http://www.mssanz.org.au/modsim09/D11/shang.pdf

See Also

ftsmweightselect, fplsrweightselect, dynupdate

Examples

dynupdateweightselect(data = ElNino, method = "pls", p = 4, errortype = "mse")

[Package ftsa version 1.3 Index]