| OCR Text |
Show THE CONTROL OF SAMPLING RELIABILITY: AN AGES OLD CONCERN Ever since man has dug the ground for useful minerals (which brings us back to antiquity), the concern for prediction of the mineral contents of potential ore from partial samples has been present. In the modern era, sampling formulas aimed at predicting the minimum amount of mineral to be taken for the sample to be representative go as far back as the end of the 19th century. Since then, with advances in statistics, applied mathematics and probability calculus, a series of theories (which have in fact been different versions of much the same theoretical basis) have been developed with unequal practical successes, until 30 years ago, when French engineer Pierre Gy placed them in a better, more complete formal framework, and developed methodologies for their application. The modern Theory of Sampling was born. Much like Geostatistics at the same period, Sampling Theory as we now know it was born from a strong need. The many publications that preceded Gy's work, especially in coal and gold mining, along with the abundant literature on published sampling standards, all testify of the perceived importance of the domain in the various fields of mining, from exploration to product sales. Yet, the number of mining companies actually using sampling theory on a routine basis is small, mostly due to the lack of clarity of the existing textbooks, and a number of difficulties in its practical implementation. A number of efforts have been undertaken more recently to correct this unfortunate situation, and the benefits of Sampling Theory have become more readily available to the mining industry than ever before. THE UNIQUE CHARACTERISTIC OF A SAMPLE AS A PRODUCT The importance of an a priori control of sample reliability stems from a unique characteristic a sample possesses as the product of a sampling procedure . Most of the procedures we apply in mining activities provide us with results, or products, which can be judged, in terms of their quality, on the basis of direct examination: for instance simple visual inspection tells us if a batch of sieved material has been sieved properly to the expected sizes, or if a model we fit to experimental data fits the data very well or poorly. A sample presents a very different case. Its only quality is what we call "representativeness" or more simply "reliability". If the sample is "representative", then the decisions based on its measurement (e.g . grade, moisture content, proportion of a given fraction size, etc.) will be well informed, and therefore taken at a low risk level; if it is not, then wrong or sub-optimal decisions will be made, with a number of unfavorable economic consequences which mayor may not be detectable and predictable. The main point is that the quality of the sample cannot be checked once the sample has been taken (say, by direct observation). The sample itself does not contain any indicators of its quality or absence thereof. In other words, once a sample has been taken, it will be used for measurements and subsequent decisions, but, unless its gathering and processing have been properly controlled before it was taken, it will always be too late , after the fact, to know whether the sample was representative. Once the sample is taken, its reliability cannot be tested. Unfortunately, our most major decisions are based on measurements made on samples. It is therefore of the utmost importance that the reliability of the samples, and the processes to which they are subjected to be turned into assay values , be controlled before they are taken, in terms of both accuraucy and precision. This is the field of application of Sampling Theory. THE UNSEEN BENEFITS OF GOOD SAMPLING PRACTICES Good sampling is obviously very desirable and often of great economic importance, but the negative impact of poor sampling is often invisible. Although specific geostatistical tools can now be |
