It was a rather "of the top of my head" answer,
but IF the Rfree reflects true error (and the SigmaA values, Cruikshank
Fit etc all use it as the basis for error estimation) then you would
expect a higher value when the data is more limited and the model less
well refined.
And IF the Rfactor at low resolution is likely to be overfitted by
having a low parameter - observation ration, which I think is often
true) then it follows that there will be a much bigger difference
between the 2 at 2.7A than at 1.7 A say...
There are many cases where this does not appear. High solvent content or
NCS can effectively increase the obsn/parameter ration.
The model structure may fit very closely to the new data and have been
refined at higher resolution
etc etc
Eleanor
John Tanner wrote:
Eleanor,
Can you elaborate on your statement about the expected value of (Rfree-R)
for moderate/low resolution structures?
Thanks.
Jack
John J. Tanner
Associate Professor of Chemistry and Biochemistry
University of Missouri-Columbia
125 Chemistry Building
601 S. College Ave.
Columbia, MO 65211
http://www.chem.missouri.edu/TannerGroup/tanner.html
begin:vcard
fn:Eleanor Dodson
n:Dodson;Eleanor
email;internet:[EMAIL PROTECTED]
tel;work:+44 (0) 1904 328259
tel;fax:+44 (0) 1904 328266
tel;home:+44 (0) 1904 424449
version:2.1
end:vcard
