Least-squares absolute-structure refinement. A case study of the effect of absorption correction, data region, stability constant and neglect of light atoms

X-ray diffraction intensity data measured on two four-circle diffractometers from a crystal of the compound potassium hydrogen (2R, 3R)-tartrate have been used to study the effect of absorption correction, data region, stability constant and neglect of light atoms in the atomic model on the least-squares estimate of the absolute-structure (inversion-twin) parameter x. It is shown that for a complete data set with Friedel pairs measured in the same plane of diffraction, an absorption correction reduced the e.s.d. of x without significantly altering its value. For data sets where only one member of each Friedel pair is allowed to contribute in the least-squares analysis, e.s.d.'s and deviations of x from its ideal value are increased compared with complete data sets containing both members of each Friedel pair. Certain aspects of combining a complete data set consisting of one asymmetric unit of reciprocal space with all-sphere measurements of absolute-structure-sensitive reflections have also been investigated. Lowering the value of the stability constant increases the e.s.d. of x, produces values of x closer to its ideal value for complete data sets and increases the variability of x for incomplete data sets. Neglect of hydrogen atoms from the atomic model produces a very small but quantifiable shift in the estimated value of x.