1
$\begingroup$

enter image description hereI'm using the R function auto.arima to fit an arima model for a time series, the result is an ARIMA(2,1,1). After that I apply the forecast function to predict some futur values. My question is Should I do the transformation ("un-differentiate" the predicted values) or is it done by forecast automatically ? edit : above is the plot I get when I execute the following code :

arimaf = auto.arima(timeseries)
pred = forecast(arimaf, h = 10)
plot(pred, main = "PREDICTION USING ARIMA(2,1,1)")
$\endgroup$
  • $\begingroup$ It's done automatically. Plot the actual values and the forecasts, and you will see that the latter nicely extend the former, there should be no big changes in the behaviour of the time series. $\endgroup$ – Richard Hardy Jun 28 '16 at 8:10
  • $\begingroup$ @RichardHardy Here is a plot of the time series + the forcast $\endgroup$ – Mohamed Nidabdella Jun 28 '16 at 8:16
  • $\begingroup$ OK, the latter does not nicely extend the former as it cannot capture all the wiggliness. But the answer is the same in any case: you do not need to manually undifference. $\endgroup$ – Richard Hardy Jun 28 '16 at 8:33
  • $\begingroup$ Your model is probably deficient as unusual values may have distorted the identification process. If you wish you can post our data and I will try and help you ( and others) as it relates to robust ARIMA model identification. $\endgroup$ – IrishStat Jun 28 '16 at 11:42

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.