Ideally, the goodness of fit (or standard deviation of an observation
of unit weight) should be as close to 1.0 as possible. Values
that deviate significantly from 1.0 may be an indication of one of the
following: (a) the model is incorrect or incomplete or
has been inadequately developed to account for disorder or solvent;
(b) the reflection data are poor or weak; (c) there is twinning
which has not been allowed for where overlap from the second twin
domain (which may have been ignored in the data collection) causes
errors in the intensities of some reflections; (d) the absorption
corrections are inadequate; (e) the weighting scheme is inappropriate.
Users of SHELXL should normally obtain goodness of fit values very close to 1.0 if none of the above features is relevant. If this is not the case, check that the weighting scheme co-efficients recommended by the program have been updated immediately prior to the last refinement run. If only the weights proposed when the model was in an early stage of development are used and not subsequently updated, the goodness of fit might deviate significantly from 1.0 and the weights should be updated.
Users of other refinement software, particularly older programs or with refinement on F, might routinely obtain values for the goodness of fit in the range 1.5-2.5. This is acceptable, but authors need to be satisfied that this range for the goodness of fit is normal for good structures determined in their laboratory. However, values for the goodness of fit that are significantly smaller than 1.0 are usually a symptom of other problems with the data or model.
|You are using an "old-fashioned", but still legal data name: _refine_ls_goodness_of_fit_obs. Please update the item to match the currently preferred name of _refine_ls_goodness_of_fit_ref. To ease this problem for the future, it is recommended that you upgrade or modify your CIF generating software accordingly.|
Copyright © International Union of Crystallography