The potential for bias in Cohen's ecological analysis of lung cancer and residential radon.