view article

Figure 3
Both panels show a type of `corrgram' (Friendly, 2002BB14) providing a rapid visual indication of the sign and magnitude of correlations between columns of the Jacobian matrix used in the final step of nonlinear least-squares refinement. The Jacobian represents the sensitivity of the prediction formula to parameters of the model. Parameters are listed in order along the rows and columns of a grid corresponding to a square correlation matrix, with parameter names as defined in Table 1[link]. The large blue circles seen down the leading diagonal indicate the perfect correlation between the effects of parameters with themselves, while large red circles show high anticorrelation. The colour and area of the circles between the extremes are related to the numerical value of the correlation coefficient, so that the most important effects stand out whereas low correlations fade to empty squares. The left panel was produced by dials.refine for refinement of a 1° data set recorded on a cubic polyhedrin crystal (Gildea et al., 2014BB15). The right panel results from multi-experiment joint refinement of five 1° sweeps of data consisting of a total of 16 indexed lattices with a shared beam and detector model. For clarity, only the crystal parameters of the same lattice as that treated alone on the left panel are shown. Comparison between the panels shows that the inclusion of multiple lattices significantly reduces the correlations between crystal parameters and other parameters of the model. For example, the correlation between the detector distance and the single-crystal lattice parameter is reduced from 0.995 to 0.323 by multiple lattice refinement.

Journal logoSTRUCTURAL
BIOLOGY
ISSN: 2059-7983
Volume 72| Part 4| April 2016| Pages 558-575
Follow Acta Cryst. D
Sign up for e-alerts
Follow Acta Cryst. on Twitter
Follow us on facebook
Sign up for RSS feeds