The largest inclusions in a piece of steel (Q1424682)

The authors deal with the problem of estimating the size of large inclusions from measurements made on a two-dimensional section of steel. The novelty of the proposed approach is that observations are not available on the variables whose extremes are of interest, so that inference has to be indirect. The approach nevertheless allows the authors to discuss prediction of the size of the largest inclusion within the whole batch of steel sampled and, most importantly, to assess the uncertainty attached to such predictions. The authors described both the maximum likelihood estimation of the generalized Pareto model, and Bayesian inference for it. The likelihood results indicate that the model fits the data well.