Conceptual

Fiducial Inference for Partially Identified Parameters in Instrumental Variable Models

A fiducial approach to causal parameters that are only bounded (partially identified) rather than uniquely determined: it derives a fiducial distribution for the identified-set bounds, samples it with an acceptance sampler to form point estimates and confidence intervals for the lower and upper bounds in binary instrumental-variable models, and interprets the sampler's acceptance rate as a diagnostic of whether the data are compatible with the instrumental-variable assumptions.