Msherif,

I wrote programs in 1968 to do exactly what you are reflecting upon.
That is a simple "try all combinations approach" . It was a failure
because of the impact of outiers, inliers, seasonal pulses, level
shifts and local time trends that were in the data. The simple one
step process that you defined also fails to recognize the importance
of diagnostic checking to revise the initally guessed model or
invertibility leading to either model simplification or additionall
diferencing.

In addition you totally ignored the PDQ ( seasonal components ) while
laying out the pdq for the base NON-SEASONAL ARIMA component.

Better to develop heuristics like the ones embedded on FreeFore (
http://www.autobox.com ) to intelligently select a starting model then
iterating via step-down, step-up and outlier detection schemes to
culminate in a useful model ..

Remember all models are wrong but some models are useful ! ( cited to
GEP BOX )

Regards

Dave Reilly
Automatic Forecasting Systems
(The makers of AUTOBOX)




[EMAIL PROTECTED] (=?iso-8859-1?q?mon=20sherif?=) wrote in message 
news:<[EMAIL PROTECTED]>...
> Hi,
> 
> Assuming the data generating process to be an
> ARIMA(p,1,q) with p=0,1,....50 and q=0,1,.....20.How
> can I find the order of ARIMA model using the Akaike
> information criterion.I thought to estimate ARIMA
> model for each order(eg
> ARIMA(0.1.1),ARIMA(0,1,2),...)and then choose the
> model with the lowest AIC,but it takes long time as I
> will estimate ARIMA model 50 times.I would be grateful
> if any one can help me to do so using one step.
> 
> Thx a bunch for your help
> 
> __________________________________________________
> Do You Yahoo!?
> Everything you'll ever need on one web page
> from News and Sport to Email and Music Charts
> http://uk.my.yahoo.com
> .
> .
> =================================================================
> Instructions for joining and leaving this list, remarks about the
> problem of INAPPROPRIATE MESSAGES, and archives are available at:
> .                  http://jse.stat.ncsu.edu/                    .
> =================================================================
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to