i was surprised to find that nobody has replied to
this yet ...
resolution is something that appears to be more
dependent on local customs or the phase of the
moon than on anything else
completeness, multiplicity and <I/sigma(I)> are
all important parameters; Rsym is less interesting
for one (arbitrary) rule of thumb, see:
http://alpha2.bmc.uu.se/~gerard/gmrp/gmrp.html
now that we have maximum likelihood refinement
the problem is probably less important since
reflections for which "only the indices have
been measured" (K.S. Wilson) should be
downweighted considerably; ML also makes it
possible to include shells with low
completeness in your refinement which one
might not have done in the old days (like
last year ;-)
--gerard
