* Create centered predictors;
data cars93a;
 set ma2612.cars93a;
 cd=displace;
 ch=hp;
 cr=rpm;
run;
proc standard data=cars93a out=ccars93a mean=0;
 var cd ch cr;
run;
* Create squares of centered predictors;
data ccars93a;
 set ccars93a;
 D2=DISPLACE**2;CD2=CD**2;
 H2=HP**2;CH2=CH**2;
 R2=RPM**2;CR2=CR**2;
run;
* Fit model using backward elimination;
proc reg data=ccars93a;
 model highmpg=cd ch cr cd2 ch2 cr2/selection=backward;
run;