CODEWITHSUNDEEP
hierarchytime-seriesforecasting
fionn
fionn

Reputation: 35

How to add regressors[xreg/newxreg] to a hierarchical time series forecast method[forecast.hts]?

I'd like an example/method to incorporate regressors into forecast.hts. Below seems correct, but alas is not producing a forecast.

require(hts)
data(htseg2)
ri <- runif(14)
rix <- runif(16)
htseg2y <- hts(window(allts(htseg2),start = c(1992), end=c(2005))[,8:17],htseg2$g)
htseg2x <- hts((window(allts(htseg2),start = c(1992), end=c(2005))*ri)[,8:17],htseg2$g)
htseg2nx <- hts((window(allts(htseg2),start = c(1992), end=c(2007))*rix)[,8:17],htseg2$g)

forecast.hts(htseg2y , h = 2, fmethod = c("arima"), positive = F, stepwise=F,ic=c("bic"),level=4,trace=TRUE
             ,xreg = allts(htseg2x)
             ,newxreg = allts(htseg2nx))

Upvotes: 1

Views: 762

Answers (1)

Rob Hyndman
Rob Hyndman

Reputation: 31800

Your hierarchical time series hts.mts includes nine series. Your exogenous variable hts.mtsx$y contains eight series. You also use the same exogenous series for the historical data as for the future data.

Upvotes: 1

Related Questions