I have been considering the question of nonisomorphism lately
after doing a review of Terwilliger's articles which describe
the refinement procedures in his program HEAVY.
Has anyone reading solved a problem with significant nonisomorphism?
What is considered to be a large nonisomorphism mean difference and
what is considered to be an impossible nonisomorphism mean difference?
I am working with models right now, but would like to hear about
practical constraints with modern instruments, area detectors
image plates, which might make a difference.
Also, when creating model data sets I assume that a gaussian
distribution of error on F^2 [I] based on mean I should be an
adequate idea of noise from instrumentation. Any reflections?
