Response to: Modelling paternal exposure as a negative control

We agree with Brew,et al.1 that the way we presented the paternal negative control directed acyclic graph (DAG) has the potential for confusion, and appreciate the opportunity to clarify our reasoning. Lipsitch ’s original negative control DAG includes nodes for measured and unmeasured confounders L and U, with a dashed line between L and U indicating that either may cause the other, and they may share common causes.2 The utility of paternal exposure B as a negative control depends on the extent to which paternal exposure shares common causes with maternal exposure A.

Source: International Journal of Epidemiology - April 24, 2020 Category: Epidemiology Source Type: research