On Thu, Oct 18, 2012 at 8:45 AM, Katherine Sippel wrote:

Part of the problem could lie in whether he input the original mtz as intensities or not. In older versions of phenix some intensities were dropped if they were below the cutoff and particularly for high res data the completeness would go down significantly. I know you have implemented the French and Wilson conversion in the newer versions but if he comes from a lab like my old one and is refining off of intensities that could be the culprit.

This is a good point, but it depends on the version of Phenix, the type of input data, and how the data file was processed. I checked my code for Table 1 and if the input data are intensities, the negative observations should *not* be discarded. The same goes for phenix.refine, among other programs. However, if amplitudes were used as input for Table 1, and had previously been filtered through a program that does not perform the French & Wilson treatment, they may be missing some data (most of which will probably be at high resolution). One caveat: the reflection file editor in Phenix has an option to convert intensities to amplitudes, but this will not run F&W, which is perhaps a bug... -Nat