Progressively censored sampling of rock joint traces

A number of sampling problems in geology and engineering geology involve geometric variables, and must deal with the almost pervasive biases that accompany geometric sampling. Among these biases is the fact that not all elements of the sampled population are fully observable. Some members, usually the largest, are censored. Inferences cannot ignore the censored members of the sample, because the censoring is often related to the variable being inferred—for example, the case of sampling for feature size. Inferences from samples are conceptually straightforward, and for the simple case of exponential parent distributions, mathematically tractible. Maximum likelihood and Bayesian results are given for the exponential case, and examples are drawn from joint surveys in rock mechanics.