research papers
Objective estimation of the error (σFmeas) in each structure factor (Fmeas), by a procedure such as that outlined in Acta Cryst. (1964). 17, 1327, allows the indicator α = [Σ (Δ2Fmeas/σ2Fmeas) ÷ χ2m − n,α]1/2 to be evaluated, where ΔFmeas = |Fmeas| − |Fcalc|, χ2m − n,α is given by the χ2 distribution at the α-significance level, and the model for which ΣΔ2Fmeas is minimized contains m independent Fmeas and n variables. The sensitivity of α to errors in Fmeas is examined under both real and hypothetical conditions. For the real case, the range in 0.01 for eight recently measured inorganic crystals indicates the average minimum residual error in σFmeas to be about 1%, the maximum error to be about 21%. The hypothetical case is considered by propagating several types of error to a set of 772 independent Fmeas corresponding to a model consisting of 10 independent atoms undergoing isotropic thermal vibration. The indicator 0.01 detects a systematic intensity error at a level in which the maximum intensity error in 97% of the data reaches 5%. Long term drift in the experiment is indicated with considerable sensitivity by 0.01. Systematic error as a function of scattering angle is detectable only if the σFmeas contain a component due to the suspected error.