How does SAS select the best regression model under several criteria such as aic, cp, bic and etc

[kunee26]

In proc reg command,

When we use 'selection=' option to select the best regression model based on various criteria  (e.g. aic, cp, bic, rsquare and etc..) 

How does SAS come up with a single best model when it needs to consider the criteria all together simultaneously?

I'm curious about this because I think different criteria may rank a model differently.

c.f)

proc reg data = /name of data/;

model Y = X1 -- Xn / selection = adjrsq mse aic bic cp  best = 1;

run;quit;

[PaigeMiller · Accepted solution]

The selection of the model is based upon SELECTION=ADJRSQ, so the model with highest ADJRSQ is selected. It is not based upon aic bic cp etc. but those statistics are reported.

[kunee26]

I've been searching around for this so long!

I feel like I've finally found oasis!

Thank you very much!