A Comparative Study Of Three Frequency-Distribution Models For Use In Ore Evaluation

A frequency distribution model plays a fundamental role in ore evaluation, especially for the feasibility study of new deep gold mines. The geological complexity of the reefs being mined at present in the Witwatersrand basin poses some difficulties regarding the modeling of the underlying ore-value distribution. This geological complexity has given rise to a need for more applicable distributional models. This paper places the geology of the Witwatersrand basin in perspective and draws some important conclusions from the genesis of the deposition. In addition to the three-parameter lognormal model, two more-general distributional models are proposed. These are the compound lognormal (CLD) and log-generalized inverse Gaussian (LN-GIG) distributions. A detailed, although preliminary, comparative study of these two more-general models with the three-parameter lognormal model is described, and the efficiencies of the estimates and modeling are analysed. It is found that the broad assumption that gold grades follow a three-parameter lognormal law is not always the case, and the need to test for departures from log normality should be emphasized so as to ensure the use of appropriate models.