On 10/5/09 9:46 AM, Ben Spiller wrote: > For the refinement case, do others think that Fcalcs become > anisotropic? Or that the liklihood method developed by Read and > others is superior even at late stages? FYI: In phenix.refine (or CNS) the total model structure factor is defined as: Fmodel = scale_overall * exp(-h*Uaniso_overall*ht) * (Fcalc + k_sol * exp(-B_sol*s^2) * Fmask) Pavel.