The efficacy of indices that characterize grain-size-frequency distributions was evaluated for response of size distribution measures to variations of a theoretical model of two normal subpopulations. Empirical data patterns of Moiola and Weiser (1968) served as a check on the reasonableness of theoretical assumptions.
Results indicate that, even for relatively simple models, percentile measures are not faithful counterparts of formal central-moment measures, except for nearly normal distributions. The percentile measures possess inherent flaws which severely compromise their utility. Data indicate, however, that combinations of two or more nonpercentile-based measures of size-frequency distributions can limit greatly the number of possible distributions for sets of statistical values and, in some cases, can define them uniquely.
Empirical biparametric scatter diagrams on the basis of percentile statistics generally conform to the theoretically produced percentile-based curves, implying a degree of natural validity of the theoretical model and also a new means of evaluating such data. Evaluation of empirical data using knowledge of theoretical results indicates that water- and wind-deposited sediment can be described effectively using relatively simple models. Much of the scatter on these plots, however, can arise as likely from sample-sediment interactions as from inherent sediment variation.