Some Advances in Lognormal Theory

This study deals with the development of some new theory concerning the bias which will result from the employment of the t-estimator if in fact the underlying distribution of ore values is not lognormal. It is found that the t-estimator is robust as long as the logarithmic variance is small and the shape factors of the transformed distribution are not unduly large. The choice of the threshold parameter [t] for the t-estimator is important, and some hints are given how to estimate it if the sample size is small, as en- countered in borehole valuation problems.