Seasonal Unit Root Tests not Produced in Forecast Server (and ETS)

[andreas_zaras]

Hello,

I have a weekly time series comprising of around 350 observations (so around 7 years). The time series exhibits seasonality and the two best models selected by Forecast Server are an Additive Winters Model and an ARIMA (1,0,1) (1,1,1) 52 whcih are both seasonal. I go to project forecasting settings and in the diagnostics tab i check the perform seasonality test. When i select the series view (in both models) anf i go to the Tables drop down menu other options are active (e.g. unit root test) but the seasonal unit root tests ooption is inactive. What i want to see is the Dickey Fuller Hasza test. Why the seasonal unot root tests option is inactive? How can i see the seasonal unit root tests? (In ETS using the same data set the Dickey Fuller Hasza test (probabilities) is not produced either).

Thnaks in advance,

Andreas

P.S. The only information that i could find concerning seasonality tests is  found as follows: When i go to the forecasing view i press the button view diagnostics nad i get a table entitled Seasonal Dummy Unit Root test summary that informs me that the seasonality is 52. There are also three cells entitled mean zeromean and trend (i guess they correspond to the three cases of the Seasonal Unit root test). In the cell entitled mean there is just a Yes value but nothing else.

[udo_sas · Accepted solution]

Hello -

From documentation:

In the Series View, the following options are now available from the Plots and Tables menus:

• Unit Root Tests provides Augmented Dickey-Fuller unit root tests for stationarity of the dependent series. The zero mean, single mean, and trend versions of the test are performed and unit root probabilities are shown at various lags using tables or bar charts.
• Seasonal Unit Root Tests are available if the series is seasonal with a seasonal cycle of up to 12 periods.

Additional details can be found here: http://support.sas.com/documentation/cdl/en/hpfug/63959/HTML/default/viewer.htm#hpfug_hpfdiag_sect02...

Thanks!

Udo