Terrible anomalous data quality when MERGE_ANOMALOUS OFF during postrefinement

[biochem-fan]

I run REFINE_METROLOGY with MERGE_ANOMALOUS ON and OFF, then MERGE with MERGE_ANOMALOUS ON.

For 2000 images (SAD),

ON: 19.65 sigma
OFF: no signal at all!

CC1/2 and the number of merged lattices were comparable.

[helenginn]

This will be addressed once processing is written to HDF5 instead of MTZ files.

^ this is no longer true. I am fixing these bugs during MTZ format.

[helenginn]

One issue has been addressed already: CORRELATION_REJECTION toggle was not honoured when doing the final merge. c58b5e4
Issues must be arising from several parameters per reflection which are not kept and are recalculated upon re-merging. My suspects are the rejection status of a Miller index and the scale factors.

Observations from merging 100 CPV images:

• If rejection criteria are set to OFF in all circumstances, the multiplicity is exactly the same between each merge.
• However, the R splits and CC 1/2s are not the same. The Rsplit and CC1/2 calculations appear to come from a smaller subset of reflections. (For last merge of post-refinement: 4178 reflections included in calculation, then 3990 reflections on final merge.) Need to see why there are fewer acceptable reflections for calculation.

More to come...

[helenginn]

This pathology should be rechecked again with OLD_MERGE OFF.