Objectives Causal reasoning should have an explicit formal structure.
Method Such a structure can be provided with the help of counterfactuals. This approach allocates different versions (factual and non-factual) of exposures and responses to every basic study unit (e.g., a subject observed at one point of time). Comparisons of these versions within the unit imply causal statements about the effect of exposures. This approach may appear unusual and strange but it is consistent to basic principles of modern physics (superposition principle of quantum mechanics).
Results The outline of causality in counterfactual terms is helpful to solve problems like defining and measuring direct and indirect causal paths or to specify biases and adjusting procedures. In contrast to experimental research observational studies (like those performed in epidemiology) suffer from missing randomization. A causal concept is important to understand the reliability of such studies: a strict counterfactual framework motivates to analyse observational studies in terms of generalised treatments (“G”). G-estimation is a procedure that defines the causal effect estimates on the individual level by counterfactual failure times. Causal models are nested within estimating models (“structurally nested failure time models”).
Conclusions Such a strict counterfactual reasoning challenges standard estimators and estimating procedures usually applied in epidemiology.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.