1. Counterfactuals and Causal Inference
2. One can use an alternative scaling such as dividing by the standard deviation of Yi (0) to eliminate this baseline model and still retain comparability.
3. It is important to note that simply varying the specification of a statistical model is not a good substitute for a formal sensitivity analysis. This might gauge how small perturbations to the statistical model may change estimates, but it does not provide any formal quantification of our uncertainty due to bias.
4. And in some cases, these categories overlap in the sense that one method can be rewritten as a special case of another.
5. Conditional independence in statistical theory;Dawid;Journal of the Royal Statistical Society. Series B (Methodological),1979