"Translate back" seasonal adjusted data into original values

[ed.sc]

Hello,

here is my question: I got sales data from a company. The data is highly seasonal and I "only" have about 100 observations, so I decided to build a forecast model for ARIMA X12 seasonal adjusted data. I made some forecasts and I would like to figure out whether my model performs better than the "model" used by the company. But: The company uses original data (not seasonal adjusted) for their forecasts. What I need to do is to "translate back" my forecasted values into the non seasonal adjusted values to get a comparison which forecast is better.

So, I seek for an opportunity to bring seasonal adjusted data forecasts back to the original form without seasonal adjustment

Is there any way to do that? I am looking foreward to your answers! Many Thanks!

Kai

[ed.sc]

Hello again,

unfortunately nobody answered my post, but I figured out the solution by myself. I would like to share the solution for other people having the same question. Here is what I figured out:

There is a command: seriesname.x12(.), which performs seasonal adjustment with ARIMA X12 as I wrote in my post above. What EViews does by seasonally adjusting a series is to decompose a series into several elements such as a cyclical component (_tc), an irregular component (_ir), a series with seasonal adjustment factors (_sf) and of cause the final seasonal adjusted series (_sa). Independent of the adjustment technique (additive, or multiplikative) one can "translate back" the adjusted series to the unadjusted values if one has all the components in the workfile. There is an option which is displayed in the following that makes EViews to safe all components in the workfile:

Code: Select all

seriesname.x12(mode=m, save="d10 d11 d12 d13" ) seriesname

where d10 gives you _sf series; d11 gives you the _sa series; d12 gives you the _tc and d13 the _ir series.

The following equtions hold for translating back series and vice versa :

Code: Select all

Additive Seasonal Adjustment: seriesname_sa = seriesname + seriesname_sf Multiplicative Seasonal Adjustment: seriesname_sa * seriesname_sf = seriesname

I hope this will help somebody in the future having the same question.

Responses are appreciated :D ,

K