I have been running regressions under the pretense that the colon acts as a 'root call.' However when I attempted to verify this assumption I found that my programs were not equivalently modelling the data. I have included sample program code and their respective results. Does anyone have any ideas? Am I accidentally programming a canonical analysis?
PROGRAM A:
proc reg;
model risk_:=male age_cat ...;
run; RESULTS: risk_radon: R2=.5239; F=1.10
risk_lead: R2=.6433; F=1.80
PROGRAM B:
proc reg;
model risk_radon=male age_cat ...;
run;
proc reg;
model risk_lead=male age_cat ...;
run; RESULTS: risk_radon: R2=.5930; F=1.26
risk_lead: R2=.6821; F=1.86
When you have multiple model statements in one REG, SAS uses listwise deletion on the variables. If you have different missing data patterns in the two outcome variables, you will get different results from the programs A and B.
Doc Muhlbaier
Duke
Ready to join fellow brilliant minds for the SAS Hackathon?
Build your skills. Make connections. Enjoy creative freedom. Maybe change the world. Registration is now open through August 30th. Visit the SAS Hackathon homepage.