Investigating the Reproducibility and Generality of the Pilot Environmental Performance Index (EPI 2006)

Abstract

The reproducibility and generality of the Pilot Environmental Performance Index (EPI) 2006 is investigated. Initially, the most accurate means of deriving the Pilot EPI 2006 scores and rankings of the countries which participated in the creation of the index from the various constructs of the Pilot EPI 2006 dataset is identified using a variety of traditional and computational intelligence tools. Use-all and leave-one-out cross-validation are subsequently employed for recreating the index values of the participating countries from the entire, as well as from parts only of the, dataset; parametric as well as non-parametric methodologies are used to this end, consequently establishing the accuracy with which the Pilot EPI 2006 scores and rankings of participating countries can be predicted from these relationships. The optimal means (combination of traditional and computational intelligence methodologies) of estimating the Pilot ESI 2006 scores and rankings of the participating countries is tested and verified; the results raise some questions concerning the reproducibility and generality of the Pilot EPI 2006.