Back in the days when a slide-rule constituted state-of-the-art computation, the Warren-Averbach method was developed to differentiate crystallite size and microstrain broadening contributions to powder diffraction profiles. The gist of this method was that one could measure the half-width of peaks directly from the powder diffraction pattern. (Back then, most were recorded by a chart recorder directly onto rolls of graph paper, so this would be a manual task.) The values were then plotted in such a way that the slope of the width function provides a microstrain estimate and the intercept provides an estimate of the crystallite size. Note that these values still needed to be adjusted to account for instrumental contributions. I have seen discussions where the Warren-Averbach is elevated as the classical and fundamental sample broadening characterization methods that all others should be validated against. I do not agree. Its real value was that the simplicity of the computations was such that it could be done in the days prior to the availability of a pocket calculator.
I feel that the Warren-Averbach approach has long been superseded by profile fitting. The sample characterization mechanism provided in GSAS-II, where peak widths are generated from sample broadening terms, as well as previously-characterized instrumental contributions, is far more accurate, in that instrumental contributions are directly accounted for and offers more advanced models, in that anisotropic sample broadening can be considered.