Urgent: Seasonal adjustment (EV 6)

[nadja123]

Hi there,

I have quarterly series with 39 observations (gross domestic product, government spending, etc.) that display seasonal pattern (zig zag). I tried Proc/Seasonal adjustment/Census X12 and Moving Average methods (additive), yielding basically the same result.

I am not sure about the following:
1. Are 39 observations sufficient to run s.a. in EV? There are some warnings in the helpfile. Is the s.a. not too aggressive if the # obs. is rather small? Is there any test on that? Under Census X12 there is Diagnostic Tab, but runs some tests only for >= 50 obs...
2. Shall I test on unit roots after or before the seasonal adjusment?
3. Can I then work with the seas. adjusted series as normally, e.g. take logs or first differences if my model requires that?

I would be extremely grateful for helping, thanks!!!!