[Gretl-users] Some questions about X-12-ARIMA

John C Frain frainj at gmail.com
Mon Jan 10 10:58:22 EST 2011

Please see below.

2011/1/10 不提供 不提供 <dear.sam at livemail.tw>:
> Dear all:
> Allin Cottrell, thank you.
> When I choose the options of Model/Time series/ARIMA/Using X-12-ARIMA to run
> the X-12-ARIMA model,
> are the explanatory variables such as as number of working days,day of week
> effects, length of month etc  inclued already?

no.  You are using the X12-ARIMA program to estimate the exact same
model as you would have estimated in "native-Gretl".  A reg-arima
model is simply an extension of the arima model in which the
disturbances in a regression are ARIMA.

 X12-ARIMA itself allows one to estimate a reg-arima model.  In
X12-ARIMA you can include certain predefined variables which are
useful in the analysis of season effects.  You can also include other
explanatory variables.  In the X12-ARIMA manual see the definitions of
the ARIMA, REGRESSION and ESTIMATE specs for details.  As I have
already stated you can use the seasonal adjustment routines in Gretl
to generate starting specs for X12-ARIMA.  This includes the data in a
format that is suitable for input to X12-ARIMA. (If the Regression
spec is not included in the spec file X12-ARIMA estimates a standard
ARIMA model).

If you wish to use the pre-defined variables in Gretl you must first
generate them before you call the  native Gretl routine.  If you call
the estimation routines from Gretl I presume that Gretl generates the
ARIMA REGRESSION and ESTIMATION options in the spec file and passes
this to X12-ARIMA.  It will not include any of the predefined optional
series (trading day, etc.) that are included in X12-ARIMA unless you
have generated them in Gretl and included them in your Gretl

In your example you have d=1 in the examples that give the same result
in both programs.  In the bad example you have d=2 and you may have
over-differencing.  While this is not always a problem it could be
lead to some instability in your estimation procedures.  I suspect
that if you forecast with both models there will probably be little
difference in your two forecasts.

Again I am not sure what you are trying to do. (seasonal adjustment,
forecasting or modeling) If you have access to Eviews you might also
try your estimates there.  Eviews uses (or used to use) different
algorithms and will probably give different answers again.

Best Regards


> If they are included automatically,
> maybe there should be differences between native-gretl and X-12-ARIMA.
> This question still puzzles me?
> Thanks a lot
> _______________________________________________
> Gretl-users mailing list
> Gretl-users at lists.wfu.edu
> http://lists.wfu.edu/mailman/listinfo/gretl-users

John C Frain
Economics Department
Trinity College Dublin
Dublin 2
mailto:frainj at tcd.ie
mailto:frainj at gmail.com

More information about the Gretl-users mailing list