* PS6.6, using DATA3-3, for Application Section 6.7 
TIME 1960
SAMPLE 1960 1993
READ(data3-3) YEAR PATENTS RD
*  Original simple regression model in Section 3.11
OLS PATENTS RD
*
* The GENR command is used to create four lagged variables (RD(t-1), RD(t-2),
* RD(t-3), and RD(t-4)) as follows:
*
GENR RD1=LAG(RD)
GENR RD2=LAG(RD,2)
GENR RD3=LAG(RD,3)
GENR RD4=LAG(RD,4)
*
* Next, the independent variables are squared.
*
GENR RDS=RD**2
GENR RD1S=RD1**2
GENR RD2S=RD2**2
GENR RD3S=RD3**2
GENR RD4S=RD4**2
*
* The SAMPLE period must be changed to begin at 1964 since the variable RD
* was lagged four time periods.  The first four observations are lost in
* this case.
*
SAMPLE 1964 1993
*
* The most general unrestricted model is estimated
*
OLS PATENTS RD RD1 RD2 RD3 RD4 RDS RD1S RD2S RD3S RD4S
*  omit variables with p-values above 0.9
OLS PATENTS RD1 RD3 RD4 RDS RD1S RD2S RD4S
*  omit variables with p-values above 0.5
OLS PATENTS RD1 RD3 RDS RD4S
*  omit one at a time 
OLS PATENTS RD1 RD3 RD4S
OLS PATENTS RD3 RD4S / PREDICT=PRED RESID=ERROR
*  COMPUTE ABSOLUTE PERCENT ERRORS
GENR ABSPCERR=100*ABS(ERROR)/PATENTS
PRINT RD PATENTS PRED ERROR ABSPCERR
DELETE / ALL
STOP