intervals: Prediction intervals

In trnnick/TStools: Time Series Analysis Tools and Functions

Description

These function return prediction intervals based on the estimated "smooth" model.

Usage

    intervals.sd(model, level=0.95, centre=TRUE)
    intervals.mae(model, level=0.95, centre=TRUE)
    intervals.mad(model, level=0.95, centre=TRUE)
    intervals.ts(model, level=0.95, centre=TRUE)
    intervals.hm(model, level=0.95, centre=TRUE)
    intervals.qr(model, level=0.95)
    intervals.qr2(model, level=0.95)
    intervals.kde(model, level=0.95, type=c("multi","single"),
                          kdetype=c("diffusion","SJ","nrd0"))
    intervals.empir(model, level=0.95, type=c("multi","single"), quantiletype=7)
    intervals.lowess(model, level=0.95, type=c("multi","single"), quantiletype=7)

Arguments

model	Object of the class 'smooth'.
level	Confidence level for the intervals construction.
centre	Defines whether the centering is needed for errors. If FALSE, then the errors are assumed to have zero mean.
type	Type of multi-step interval calculation: ‘multi’ for using multiple steps ahead errors and ‘single’ for using one step ahead error only.
kdetype	Type of kernel used for intervals.kde. See kdemode for details.
quantiletype	Quantile algorithm used. See quantile for details.

Details

• intervals.sd() - intervals produced using sigma values for each horizon;

• intervals.mae() - intervals produced using MAE (Mean Absolute Error) for each horizon;

• intervals.mad() - intervals produced using MAD (Median Absolute Deviation) for each horizon;

• intervals.ts() - intervals produced using two sigma values;

• intervals.hm() - intervals produced using half moment;

• intervals.qr() - intervals produced using quantile regressions Taylor and Bunn (1999) style;

• intervals.qr2() - intervals produced using quantile regressions based on power function;

• intervals.kde() - intervals produced using kernel density estimation;

• intervals.empir() - intervals produced using empirical distribution;

• intervals.lowess() - intervals produced using lowess smoothing of edf;

Value

lower	Lower bound of prediction interval.
upper	Upper bound of prediction interval.
sigma	Matrix with lower and upper sigmas.
hmValues	Vector with values of half moments for each horizon.

Author(s)

Ivan Svetunkov and Nikolaos Kourentzes

References

Nothing here yet.

See Also

Nothing

Examples

    library(smooth)

    ourModel <- es(rnorm(100,100,10))
    intervals.sd(ourModel)
    intervals.mae(ourModel)
    intervals.mad(ourModel)
    intervals.ts(ourModel)
    intervals.hm(ourModel)
    intervals.qr(ourModel)
    intervals.qr2(ourModel)
    intervals.kde(ourModel)
    intervals.empir(ourModel)
    intervals.lowess(ourModel)

trnnick/TStools documentation built on Sept. 14, 2019, 5:22 a.m.