In causal mediation analysis, natural effects are identified only under strict assumptions involving cross-world counterfactuals. An alternative approach recently developed, called separable, allows for identification of mediational effects in a wide range of models, since it relies on weaker assumptions than those required by natural effects. In this paper, the separable-effect approach is revised and an application to data is presented.
|Titolo della pubblicazione ospite||Book of Short Papers - SIS 2021|
|Numero di pagine||6|
|Stato di pubblicazione||Published - 2021|