The importance of specification uncertainty and intolerance to measurement error in a study of the impact of dentine lead on childrens’ IQs

Needleman et al. concluded that, on average, an increase in the dentine lead of children from <6 ppm to >24 ppm was associated with a 4‐point IQ decrement. Recently, a spate of papers has conjectured that the reports of Needleman et al. and others could be spurious because of measurement error and omitted variable bias. In this Bayesian analysis of the reconstituted Needleman et al. data set, we replace such conjectures by empirical tests. Even though we structure the tests to provide very favourable opportunities for the conjectures to acquire support, the possibility is seen to be remote that omitted variable bias significantly alters the estimated influence of lead exposures upon childrens’ psychometric intelligence. Similarly, the Bayesian analysis demonstrates that other included covariates are likely to affect the estimated influence of lead exposures only if they are quite poorly measured.

Key words::

• Dentine lead
• childrens’ IQs
• Bayesian diagnostics