Answers are grounded in this paper and cite their sources.
Unmeasured confounding poses a common problem in observational studies. Although sensitivity analysis based on marginal structural models can now be used in the sensitivity analysis of the causal risk difference, this method cannot be simply applied to the causal risk ratio. Here, we propose a method of sensitivity analysis for the causal risk ratio that extends previous work on the causal risk difference, and apply it to a classic cohort study investigating behavior patterns in coronary heart disease. We also extend the proposed method to a sensitivity analysis of the causal risk ratio of a time-varying treatment on repeated outcomes. The proposed method helps to provide a realistic picture of the potential impact of unmeasured confounding.
No takes yet. Share an insight, caveat, or question.
Yasutaka Chiba (2009) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: