Estimating uncertainty in observational studies of associations between continuous variables: example of methylmercury and neuropsychological testing in children