As has been shown recently (and discussed on this board), Rsym is not the best measure of data quality (if any measure at all):
http://www.sciencemag.org/content/336/6084/1030.abstract > narayan viswam wrote: >> Hello CCP4ers, >> In my data, the highest reolution shell 2.8-3.0 A has I/sigmaI 2.5 & >> Rsym 224.3 % >> for multiplicity 7.8 and completeness 98.2 %. I solved the structure by >> MAD & refined it >> to Rfree 27.3 %. Ths crystal belongs to P622 space group and it is not >> twinned. The water >> content is 68%. I loweredthe multiplicity to 4.1 by excluding few >> images but still the >> Rsym is > 200 % and I/sigmaI > 2.0. My rudimentary crystallography >> knowledge makes me >> believe it's quite Ok to use data upto 2.8 A and report the statistics. >> Could I request >> people's views. Thanks very much. >> Narayan > > After refinement, what is R-free in the last shell? If it is significantly > better > than random, say around .4 or less, that could be taken as evidence that > there > is data in the last shell. > Also check the error model- Rsym >2 sort of implies the error is greater > than > the signal, so I/sigI 2 seems surprising. > eab > -- Edwin Pozharski, PhD University of Maryland, Baltimore