Origin and characterization of departures from the statistical model of component-peak overlap in chromatography.

A statistical model of component-peak overlap in complex chromatograms is reviewed. Procedures for the estimation of the number of components in an analyte from its chromatograms by means of this model are restated. We note that the statistical model does not account for the effects of certain realistic chromatographic attributes. The influences of component-peak density, amplitude range, asymmetry, and noise levels on the estimation of the average component number are determined empirically with computer-generated chromatograms and are quantified by analyses of variance. We find that small departures from the model arise from variations in the magnitude of the amplitude range, density and the noise level. A large departure from theory arises from an application of the model to chromatograms containing highly asymmetric component-peaks. In spite of these departures, the estimation of the component number from chromatograms containing randomly distributed Gaussian component-peaks is uniformly more accurate with the use of the model than from a counting of peak maximum in chromatograms of extraordinarily high resolving power.