Expert knowledge, confounding and causal methods.
Gaceta sanitaria / S.E.S.P.A.S, (December 2001)2767<m:linebreak></m:linebreak>Causalitat.Abstract
The absence of unmeasured confounding is the fundamental condition for causal inference from observational data. Even when this condition holds and the models are correctly specified, estimates from standard methods may not have a causal interpretation if there are time-dependent confounders that are affected by prior exposure. Causal methods, such as marginal structural models and structural nested models, avoid this problem.