The least-squares goodness-of-fit parameter S for
               significantly intense reflections (see
               _reflns.threshold_expression) after the final cycle of
               refinement. Ideally, account should be taken of parameters
               restrained in the least-squares refinement. See also
               _refine.ls_restrained_S_ definitions.
                   {  sum { w [ Y(obs) - Y(calc) ]2 }  }1/2
               S = { ----------------------------------- }
                   {            Nref - Nparam            }
               Y(obs)  = the observed coefficients
                         (see _refine_ls_structure_factor_coef)
               Y(calc) = the calculated coefficients
                         (see _refine_ls_structure_factor_coef)
               w       = the least-squares reflection weight
                         [1/(u2)]
               u       = standard uncertainty
               Nref   = the number of reflections used in the refinement
               Nparam = the number of refined parameters
               and the sum is taken over the specified reflections