Compare admissible-set estimates with a scientifically meaningful target

Determine how to compare the proximity of estimates induced by different admissible trajectory sets, including sets corresponding to different grace-period lengths, to a scientifically meaningful target parameter.

Background

The paper formalizes clone-censor-weight estimation using admissible trajectory sets and shows that changing the admissibility definition, such as by changing a grace period, changes the associated stochastic intervention and therefore the causal estimand. The discussion proposes sensitivity analyses over multiple admissible sets but identifies as unresolved how the resulting estimates should be evaluated against a scientifically meaningful target parameter.