Faster calculation of the full matrix for least-squares refinement

Equations derived from a statistical model and valid for all space groups give estimates of the elements of the matrix for least-squares refinement of atomic coordinates and isotropic thermal parameters for a large crystal structure with many diffraction data. The elements are functions of the lengths and directions of Patterson vectors and the distribution of weights as a function of Bragg angle. For a large data set, this matrix, or a portion of it, can be calculated in a time that is approximately proportional to the number of elements calculated and independent of the number of reflections.