FREE_R set

Hong Ling hong at MYCROFT.MMID.UALBERTA.CA
Fri Nov 13 16:17:07 EST 1998


Hi,

The purpose of Rfree is to monitor any overfitting. At the  stage of 
rigid body refinement there is no any chance for refinement to 
overfit model (much fewer parameters that reflections). Therefore, it is 
possible that Rfree is higher than Rwork in rigid body refinements and the 
beginning of positional refinement right after. I agree with Bart Hazes's 
opinion, as long as the test data is random for certain data at certain 
resolution and uniformly distributed in the data set, the test data is 
ok, why bother to generate it differently?

Best wishes,

Hong
----------------------------------------------------------------
Hong Ling                
Dept. of Biochemistry     |E-mail: hong at mycroft.mmid.ualberta.ca
Univ. of Alberta          |Phone: 403-492-4696 (lab)
Edmonton, AB                      403-492-7521 (fax)
Canada T6G 2H7                   
----------------------------------------------------------------




More information about the X-plor mailing list