scale factors

Savvas N. Savvides savvas at ODYSSEUS.ALS.ORST.EDU
Fri Nov 13 14:13:18 EST 1998


Hi everyone,
        an issue that came up in a recent discussion here at the lab is
the fact that in refinement routines XPLOR computes two "optimal" scale
factors, one for the test set and one for the working set.  From our
experience these two scale factors usually differ by 1-3% at the end of
refinement although in a couple of cases from previous work of mine those
scale factors are identical.  The concern is about the need to have a
special scale factor for the test set, especially since it is estimated
using only 5-10% of the reflections.  Wouldn't it be more correct to
estimate a scale factor using the entire data set and then
apply this scale factor for calculating R-free and R-work?
        To make things more interesting, in a couple of tests I have done
in which I used the "optimum" scale factor for the working set to
calculate R-free, R-free is worse (by about 0.2%) than when I allow XPLOR
to use the "optimum" scale factor for the test set as it normally does. 
        Any thoughts out there on this issue? 

Best wishes to everyone,

Savvas


*****************************************************************************
*                         SAVVAS N. SAVVIDES                                *
*                                                                           *
*                      E-mail: sns2 at cornell.edu                             *
*                                                                           *
* In Absentia at:                  Currently at:                            *
* --------------                   ------------                             *
* Program in Biophysics            Department of Biochemistry & Biophysics  *
* 221 Biotechnology Building       2011 Agricultural and Life Sciences      *
* Cornell University               Oregon State University                  *
* Ithaca, NY 14853                 Corvallis, OR 97331-7305                 *
* USA                              USA                                      *
*                                                                           *
* Tel:(607)-255-8432               Tel:(541)-737-3196                       *
* Fax:(607)-255-2428               Fax:(541) 737-0481                       *
*****************************************************************************




More information about the X-plor mailing list