Editor’s Invited Article

Journal of Agricultural, Biological, and Environmental Statistics

, 8:253

First online:

A Bayesian hierarchical model for risk assessment of methylmercury

  • Brent A. CoullAffiliated withDepartment of Biostatistics, Harvard School of Public Health Email author 
  • , Maura MezzettiAffiliated withIstitutodi Metodi Quantitativi, Università Bocconi
  • , Louise M. RyanAffiliated withDepartment of Biostatistical Science, Harvard School of Public Health and Data Farber Cancer Institute

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access


This article uses a Bayesian hierarchical model to quantify the adverse health effects associated with in-utero exposure to methylmercury. By allowing for study-to-studyas well as outcome-to-outcome variability, the approach provides a useful meta-analytic tool for multi-outcome, multi-study environmental risk assessments. The analysis presented here expands on the findings of a National Academy of Sciences (NAS) committee, charged with advising the United States Environmental Protection Agency (EPA) on an appropriate approach to conducting a risk assessment for methylmercury. The NAS committee, for which the senior author (Ryan) was a committee member, reviewed the findings from several conflicting studies and reported the results from a Bayesian hierarchical model that synthesized information across several studies and for several outcomes. Although the NAS committee did not suggest that the hierarchical model be used as the actual basis for a methylmercury risk assessment, the results from the model were used to justify and support the final recommendation that the risk analysis be based on data from a study conducted in the Faroe Islands, which had found a positive association between in-utero exposure to methylmercury and impaired neurological development. We considera variety of statistical issues, but particularly sensitivity to model specification.

Key Words

Benchmark dose Dose-response Gibbs sampling Linear mixed model Meta-analysis Multiple outcomes Win BUGS