Sensitivity Analysis and Epidemiologic Triangulation

Abstract

Epidemiologic triangulation integrates evidence from a variety of studies that have unrelated sources of bias. Sensitivity analysis assesses how a conclusion would change if assumptions were relaxed. Causal interpretations of associations between measures of health and air quality require a non-confounding assumption. Exposures are not randomly assigned nor selected, but with integrated sensor monitoring systems and annotation a key confounding parameter can be estimated. This talk introduces a novel statistical methodology for causal inference based on data transport and collaboration between exposure scientists and environmental epidemiologists.