Hello All, I've noticed that the figure-of-merit (FOM) and weighted difference map coefficients (e.g. DELFWT) for the test set (FreeR_flag = 0) of reflexions in an MTZ file as produced by a certain well-known ML refinement program are all zero, while the same coefficients produced by another well-known refinement program are non-zero. I can see that there might be arguments to justify both approaches, e.g. the values for the test set might be over-estimated so having them all zero is safer, but on the other hand a non-zero value even if it's biased must still be closer to the correct value than zero is.
I was wondering if anyone knows of a reference where the pros and cons of the alternative approaches are discussed? Cheers -- Ian
