It might be instructive to draw a precession-like diagram of your reflections in reciprocal space. Remember that reciprocal space dimensions are generally in reciprocal Angstrom and volumes raise those dimensions to the third power. Thus (1/12)^3 to (1/15)^3 is not a big volume.
How many reflections should you have between 15 Angstrom and 12 Angstrom? Between 12 Angstrom and 4 Angstrom? Suppose only 4 unique reflections exist between 12 and 15 Angstrom. What happens to your completeness if you are missing just one of them? -----Original Message----- From: CCP4 bulletin board [mailto:ccp...@jiscmail.ac.uk] On Behalf Of Simon Kolstoe Sent: Thursday, June 24, 2010 10:05 AM To: CCP4BB@JISCMAIL.AC.UK Subject: [ccp4bb] Another scaling question Dear CCP4bb, I am still playing around scaling two datasets together and have noticed another interesting behavior in scala. If I scale all my data (from 1.5A to 51A) I get 100% completeness in my outer shell, 98% in my inner shell and 99.9% overall, stats that I am normally quite happy with. I tend to also look at the table in the log file which in this case reports above 98% completeness in all shells between 1.5 and 4.7A. Rmerges are 0.054 overall with 0.29 in the outer shell, which again I think is OK. However I then ran scala again in an attempt to scale with the strongest overlapping reflections in my two datasets, so limited the resolution to between 15A and 4A. Now when I look at my completeness I get 97% overall, but only 32% in the highest 15-12A shell! Is something funny going on in the program or am I really missing 70% of my data in this resolution, and if so how come the scala run with all the data doesn't report this? I am now worrying that all the data I previously thought was complete might be lacking many lower resolution reflection! Thanks, Simon