ABSTRACT
Environmental risk assessment relies on statistical methods for estimation of the
dose-response relationship and for subsequent calculation of exposure limits. Interspecies
extrapolations are avoided when human epidemiological data is available. However,
observational studies always involve concerns regarding confounding, measurement error,
missing data, and multiple comparisons. Standard statistical techniques, such as multiple
regression analysis, are poorly suited for such data and will lead to biased and inefficient
effect estimation and safe dose calculation. This paper explores the potential of structural
equation models for improving current statistical methods in environmental risk
assessment. The concreter case of neurobehavioral effects in Faroese children prenatally
exposed to methylmercury is used for illustration.
