1
$\begingroup$

I'm trying to use an ARIMA model to generate forecasts using monthly revenue data that exhibits both trend and seasonality.

I've used the Expert Modeler from the SPSS Forecasting Module to determine that an

ARIMA (3, 1, 0)(0,1, 0)

best fits my data. I've noticed that the Expert Modeler only includes an AR term at Lag 3 because the lower orders are not statistically significant.

I would like to fit a custom ARIMA that mimics the Expert one as closely as possible so that I can specify a Level Shift Outlier that the Expert Modeler does not pick up (p-value = .012). It is important that I include this outlier because it:

  1. Makes sense intuitively
  2. Improves the fit of the model drastically

However, I can't seem to find a way to specify a custom ARIMA Model that excludes the statistically insignificant lower order AR terms, or a way to manually specify outliers in the Expert Modeler.

I was wondering if there is an option to handle this in SPSS? Or if it is something that I should even be concerned about?

$\endgroup$
1
  • $\begingroup$ post a link to your data $\endgroup$
    – Tom Reilly
    Nov 23, 2015 at 19:46

1 Answer 1

1
$\begingroup$

you have to open a syntax windows and change the parameters manually.

Create and run your model with all lags. Then, copy the TSMODEL data tha your output shows before your model. Usually, it comes like this:

PREDICT THRU YEAR 2000 QUARTER 4. 
* Time Series Modeler. 
TSMODEL 
   /MODELSUMMARY  PRINT=[NONE] 
   /MODELSTATISTICS  DISPLAY=YES MODELFIT=[ SRSQUARE RSQUARE] 
   /MODELDETAILS  PRINT=[ PARAMETERS RESIDACF RESIDPACF FORECASTS]  PLOT=[ RESIDACF RESIDPACF] 
   /SERIESPLOT OBSERVED FORECAST FIT 
   /OUTPUTFILTER DISPLAY=ALLMODELS 
   /AUXILIARY  CILEVEL=95 MAXACFLAGS=24 
   /MISSING USERMISSING=EXCLUDE 
   /MODEL DEPENDENT=GDP 
      PREFIX='Model' 
   /ARIMA AR=[12,11,10,9,8,7,6,5,4,3,2,1]  DIFF=1  MA=[0]  ARSEASONAL=[0]  DIFFSEASONAL=0  MASEASONAL=[0] 
      TRANSFORM=NONE  CONSTANT=YES 
   /AUTOOUTLIER DETECT=OFF.

After that, go to FILE > NEW > SYNTAX and paste what you have copied. Now, you only have to change the AR or MA parameters:

AR=[12,11,10,9,8,7,6,5,4,3,2,1] for AR=[8,4,2,1] and etc.

$\endgroup$
1
  • $\begingroup$ This is exactly what I was looking for. Thank you! $\endgroup$
    – Rick Arko
    Dec 10, 2015 at 16:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.