Practical guidelines for conducting sensitivity analysis
As you’ve seen, sensitivity analysis is a powerful tool in causal inference for assessing the robustness of research findings under various assumptions. However, its effectiveness depends on thoughtfully selecting parameters for analysis, accurately interpreting the results, and understanding its limitations. In this section, we’ll cover some practical guidelines for conducting sensitivity analysis that can help you tackle these aspects effectively. First, we’ll learn how to choose parameters.
Choosing parameters for sensitivity analysis
Begin by identifying the key assumptions underpinning your causal model. This includes assumptions about the absence of unmeasured confounders, the type relationship between variables (for example, linearity), and the homogeneity of effects across subpopulations.
Parameter deep dive
Once you’ve done this, you must prioritize parameters for which there is...
