The Bootstrap - A Non-Parametric Statistical Method for Assessing the Reliability of Alluvial Gold Grade Estimates

Alluvial gold grades are typically highly erratic. It is very often difficult to establish with confidence the in situ grades of reserve blocks due to the small number of pit or drill hole samples available, and their erratic values. Trends in grade are often obscure. Geostatical methods are inappropriate due to the extreme nugget effect found in alluvial gold field data. Classical statistics appear more suited to such data but require assumptions to be made regarding the nature of the population sampled. Recent developments in computer intensive non-parametric statistical methods provide a new approach to dealing with some of the above problems. These methods avoid making assumptions about the type of population distribution, usually through some kind of re-sampling scheme. The bootstrap procedure involves taking the estimate of interest and synthesising its population of by re-sampling the original data a large number of times (1000+), at random and with replacement. The synthetic population is then used to draw conclusions about the likely behaviour of the true population.