Abstract
The accuracy of grouped moment statistical parameters is evaluated by comparing the grouped values to corresponding values calculated from digitally generated, ungrouped grain-size data. Errors in grouped statistical measures in general arc small, thus environmental interpretations resulting from their use are consistent with the actual characteristics of the size-weight frequency distributions of samples. Errors due to grouping data into size classes are smallest for the standard deviation and mean, and are largest for kurtosis. In general, the accuracy of the statistical parameters increases as the analysis interval decreases from 1.0phi to 0.25phi , although changes in the values of these measures resulting from such a decrease in the analysis interval in most cases are small. Truncation of sample distributions does not seriously affect the accuracy of the statistical measures, although statistics obtained from the distributions used in this study and truncated at 9phi are less accurate than those truncated at 12phi . Attempts to artificially reconstruct size fractions within the "unanalysed" portions of truncated distributions resulted in decreased accuracy of derived statistical measures. Thus unanalysed material in fine-grained sediment samples should be assigned to a single size fraction with an appropriate mid-phi interval. The accuracy of grain-size statistical parameters is not consistently improved by generalized correction factors, such as Sheppard's corrections, in part because the fine details of particle-size distributions, to which statistical measures are sensitive, are not taken into account by the corrections.
