Cochrane-Orcutt Method in SPSS

Estimation of Regression model using AR(1) errors

 

    1.     Open CORP data set

    2.     Go to Transform → Compute and create the following variables:

            a.    lprof = ln(corpprof)

            b.    lgnp = ln(gnp)

            c.    lprof1 = lprof - .748*lag(lprof)

            d.    lgnp1 = lgnp - .748*lag(lgnp)

            e.    intcpt1 = 1 - .748

    3.     Go to Analyze → Regression → Linear:

            a.    Independent = lprof1, Dependent = intcpt1 lgnp1

            b.    Click the “Options” button, and unselect the checkbox labeled “Include constant in equation”

            c.    Click the “Statistics” button, and select the “Durbin-Watson” checkbox

            d.    Save unstandardized residuals

    4.     Go to Graphs → Scatter → Simple:

            a.    res_1 vs. year

            b.    Title “Diagnosis of Autocorrelation in transformed model: Time Series Plot”

    5.     Go to Transform → Compute: Target Variable = lagr, Numeric Expression = lag(res_1)

    6.     Go to Graphs → Scatter → Simple:

            a.    res_1 vs. lagr

            b.    Title “Diagnosis of Autocorrelation in transformed model: Plot of resid vs. lagged resid

    7.     Go to Analyze → Correlate → Bivariate: Variables = res_1, lagr