Evaluation errors and value frequency distribution models of the gold ore bodies in the Witwatersrand and Free State systems are reviewed. It is found that the low-grade sections of the data are prone to the largest evaluation errors, and that an understanding of the shape of the value frequency distribution is essential to the calculation of mean grade. The evaluation of a mineral process for a particular ore requires attention to mil-feed sampling, sample processing, and monitoring. In particular, the calculation of a representative mass of mill-feed sample (Ms) should take account of size-by-size value variance. Alternative Ms models are discussed. |