Assessing the effect of clustered and biased multi-stage sampling

Abstract

We propose a method for detecting biased multi-stage sampling of spatial data and a method to adjust for biased Clustering of samples. We assess the effect of these methods for the analysis of radioactivity contamination data from Rongelap island, with the scientific problem being the estimation of the maximum level of radioactivity over the island. These data were collected over a two-stage process of uniform and clustered samples, which may have an impact on conclusions from a standard analysis that does not account for either of these features.