Nonparametric analysis of air pollution indices compositions

Abstract

Synthetic indices are useful tools used to summarize multivariate information by means of a single value. In the environmental framework these indices are frequently proposed both for measuring pollution and for measuring the associated risks. The drawback of using synthetic indices is, however, that some important information about their components is lost. In this paper we propose to save and interpret these components by transforming them into compositional data. The interpretation of compositions is performed conditionally on the meteorology. In particular, much attention is directed towards explaining the absence of specific pollutants in the index determination.