Design and analysis for modeling and predicting spatial contamination

Sampling and prediction strategies relevant at the planning stage of the cl eanup of environmental hazards are discussed. Sampling designs and models a re compared using an extensive set of data on dioxin contamination at Piazz a Road, Missouri. To meet the assumptions of the statistical model, such da ta are often transformed by taking logarithms. Predicted values may be requ ired on the untransformed scale, however, and several predictors are also c ompared. Fairly small designs turn out to be sufficient for model fitting a nd for predicting. For fitting, taking replicates ensures a positive measur ement error variance and smooths the predictor: This is strongly advised fo r standard predictors. Alternatively, we propose a predictor linear in the untransformed data, with coefficients derived from a model fitted to the lo garithms of the data. It performs well on the Piazza Road data, even with n o replication.