There does indeed
appear to be a tendancy for the SHELXC values to become slightly
negative at the high resolution end, i.e. when a value of zero would be
expected for pure noise, so maybe it is something more fundamental?!
At zero signal, wouldn't CC1/2 be
dominated by the "random" distribution of the finite number of observations over the two half-sets?
I.e. with a true average of zero, occasional higher intensities in one half-set would
correlate with lower intensities in the other (and vice versa)?
Sebastiaan.
Dr. Sebastiaan Werten
Institut für Biochemie
Universität Greifswald
Felix-Hausdorff-Str. 4
D-17487 Greifswald
Germany
Tel.: 
+49 38 34 86 44 61
e-mail: [log in to unmask]
