Statistical estimation of chemical characteristics for dispersed solid matter

Abstract

Statistical evaluation of data from chemical analyses of solid waste incinerator slag suggests that the average concentration of an element and other characteristics are inaccurately esti mated when obtained from a skewed frequency distribution. Estimation procedures for central value, dispersion measure and limits of element concentration in the total slag output of an incinerator are found using geometric properties of the lognormal distribution. Functional relationships between the coefficient of variation and expected errors in geometrically derived sample statistics are described on a theoretical level. A sampling strategy for the control of accuracy in the geomet ric mean is developed. It is shown that confidence intervals for two-sigma concentration limits can be constructed by simultaneous estimation of logmean and logvariance. Our results reveal that the expected sampling error in the geomet ric mean concentration exceeds 30% and in the upper con centration limit 80% when less than a hundred observations per element are available. It is concluded that chemical char acteristics of heterogeneous matter should rather be based on interval estimates than point estimates.