*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
Carlos Frazao wrote:
I understand that in earlier days, when data were much-much harder to collect, crystallographers were very happy with complete (although unredundant) data set. Then, Rmerge(or Rsym) were perfectly adequated in pratical terms, as data collection expriments were designed essentially to assure that at the end of the week data were complete - it was not pratical to collect redundant data. Today, with ccd's and synchrotron radiation, we quite often try to obtain data as redundant as possible, in particular to detect anomalous signal, and Rmerge statistics are simply not applicable anymore.
I just want to clarify that Rmerge or R-sym are _not_ appropriate at low redundancy- they are appropriate at high redundancy (where they approach the value of Rmeas) and at low redundancy they underestimate the deviation by a factor of up to sqrt(2)=1.41. The point is they were _used_ at low redundancy, contributing (together with 3-sigma cut-offs) to unrealistically rosy estimate of the deviation. It is inappropriate to apply expectations that were developed during those days to data which is correctly collected by todays standards. If someone says Rsym is too high in the last shell, ask what (s)he thinks Rmerge would be in a shell with no data. 100%? No, it would approach infinity, because there are equal number of positive and negative intensities and the mean intensity, which appears in the denominator, is zero. At least it is if you use a cuttoff on mesurements of -3 sigma. With a cutoff of +3 sigma this is quite different, as there are no negative intensities even where there is no data. Ed
