Sieved modern coastal sands were used to compute standard grain-size parameters from all known graphic formulas and the method of moments. Progressively greater variation existed in computation of mean, sorting, skewness, and kurtosis. As a group, Folk and Ward's parameters came closest to approximating moment parameters.

The median yielded the greatest deviation from the moment mean, being too fine for coarse-skewed samples and too coarse for fine-skewed samples. Graphic means were very close to the moment mean. Formulas (Folk and Ward, McCammon) considering points near the tails of the distribution and the median produced almost the same values as the moment mean. Formulas (Otto, Inman) omitting the median yielded slightly too coarse means for coarse-skewed samples and slightly too fine for fine-skewed samples. Sampling more than four frequency curve points did not increase accuracy of means or sorting estimates.

Among graphic sorting estimates, Krumbein's QD_^phgr consistently was appreciably lower and Friedman's SO_s and Griffiths' PD_^phgr consistently were appreciably higher than moment standard deviation. Cadigan's S and Otto's and Inman's ^sgr_^phgr produced irregular values averaging higher and lower, respectively, than moment values. Folk and Ward's ^sgr1 and McCammon's ^sgr yielded consistent results slightly below moment values. Tanner's D_3,29 produced the closest approach to moment values. The non-phi-unit methods (Trask's So; Miller and Zeigler's DS; Sharp and Fan's Si) showed the same relative change in sorting between samples.

No graphic skewness estimate correlated well with moment skewness, although they usually yielded the same sign. Only Folk and Ward's Sk₁ and Inman's ^sgr_^phgr produced fairly consistent results close to the average skewness value. No graphic kurtosis estimate correlated with moment kurtosis, which was consistently higher.

End_of_Article - Last_Page 735------------