Troubleshooting
Problem
I've run a linear regression model and found that the Durbin-Watson statistic indicates autocorrelation among the residuals. I would like to use a method such as a Cochrane-Orcutt autoregression model to handle this, but I don't see how to specify this in SPSS. Where do I find this?
Resolving The Problem
The Cochrane-Orcutt method is available in the AREG procedure. If you are running Release 14 through 18, you will only have access to the AREG procedure if you previously had the Trends module licensed and have continued to do so. In Releases 14 and 15, AREG is accessible via the menus (Analyze>Time Series>Autoregression). In later releases, it is available only through command syntax. The basic structure required is:
AREG VAR=dependent WITH independent(s)
/METHOD=CO.
In Release 19, AREG is available (still only via command syntax) to all users of IBM SPSS Statistics, with or without the Forecasting or Trends Original licenses.
If you are running a release from 14 through 18 and do not have access to the AREG procedure and want to correct for autocorrelated errors, if you have the Forecasting module, you can use the TSMODEL procedure to fit the ARIMA(1,0,0) model fitted in AREG. (However, TSMODEL uses maximum-likeihood estimation. It does not offer Cochrane-Orcutt estimation.) Specify Analyze>Forecasting>Create Models. Specify your dependent and independent variable(s). From the Method dropdown menu, choose ARIMA. Click on the Criteria button. Specify a 1 for the Nonseasonal Autoregressive structure. Click Continue. To see the parameter estimates, click on the Statistics tab, and under Statistics for Individual Models, check the box for Parameter estimates. Click OK.
Related Information
Historical Number
57678
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21481519