Diffraction intensities measurements are influenced by random errors and complex patterns of systematic effects. The systematic effects can be physically modeled if their sources are known, resulting in deconvolution of experimental data into: the signal arising from crystal structure, other signals, for instance absorption or specific radiation-induced changes, and experimental errors. The systematic effects that are not properly modeled contribute to the error estimates, effectively decreasing the, already low, phasing signal-to-noise ratio. Data processing programs, for instance Denzo and Scalepack, have built-in hierarchy that allows for optimal deconvolution of signals and errors. Their analysis relies on comparing the intensities of symmetry-equivalent reflections using multivariate statistics methods. Multicomponent modeling of variance is particularly useful for correcting the diffraction data affected by radiation damage.