Parkin S
Biochemistry Department, Duke University Medical Center, Durham, NC, USA.
Acta Crystallogr A. 2000 Mar;56 (Pt 2):157-62. doi: 10.1107/s010876739901497x.
Assessment of quality in crystal structure determination entails analysis of global statistics. In data reduction, quality is assessed using R(merge) and mean I/sigma(I). Progress in structure solution and refinement is checked by the goodness of fit, variants of the R index, R(cryst), and its cross-validation counterpart, R(free). These statistics are useful and provide a convenient means of comparison but their scalar nature renders them unable to capture the essence of three-dimensional entities such as diffraction patterns and molecular models. A simple general method to quantify spatial variations in scalar statistics has been developed. In it, a symmetric matrix, the R tensor, is used to represent the local average residual as a function of diffraction geometry. An effective value of the statistic in question can then be found for any direction in reciprocal space. Differences between these effective R indices for individual reflections or groups of reflections can help to steer refinement strategy and assess the final structure.
晶体结构测定中的质量评估需要对整体统计数据进行分析。在数据处理过程中，使用R(merge)和平均I/sigma(I)来评估质量。通过拟合优度、R指数的变体R(cryst)及其交叉验证对应物R(free)来检查结构解析和精修的进展。这些统计数据很有用，提供了一种方便的比较方法，但它们的标量性质使其无法捕捉诸如衍射图案和分子模型等三维实体的本质。已经开发出一种简单的通用方法来量化标量统计数据中的空间变化。在该方法中，使用一个对称矩阵，即R张量，来表示作为衍射几何函数的局部平均残差。然后可以在倒易空间的任何方向上找到所讨论统计量的有效值。单个反射或反射组的这些有效R指数之间的差异有助于指导精修策略并评估最终结构。
