[R] predicted values

[ripley@stats.ox.ac.uk]

On Mon, 10 Mar 2003, Luis Silva wrote:

> Hello helpers
> 
> I fitted an SARIMA model to a time series and would like to 
> predict 10 steps ahead. I made
> 
> >dataset1.arima<-arima(dataset1.ts,order=c(2,1,0),seasonal=list
> (order=c(1,1,0)))
> 
> >predict(dataset1.arima,n.ahead=10)
> 
> The problem is that the predicted values came in a different 
> scale. The original series has values like  26992411 and the 
> predicted values are like -0.2341768144. Why is this?

User error?  If we take the example in ?predict.Arima and use teh same 
orders as yours

data(USAccDeaths, package="ts")
USAccDeaths
fit <- arima(USAccDeaths, order = c(2,1,1), seasonal = list(order=c(1,1,0)))
predict(fit, n.ahead = 10)

it gives results of the expected size.  So how can we possibly debug your 
private example?

BTW, arima0 can also be used.

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595