Skip to main content

Events

Matching Estimators using Misspecified Generalized Empirical Likelihood

Matching Estimators using Misspecified Generalized Empirical Likelihood


Event details

Abstract

We cast causal parameters as estimands in a generalized empirical likelihood problem.  Even when assumptions on random or pseudo-random assignment are justified, the relevant generalized empirical likelihood problem is misspecified.  We connect the pseudo-true value of the misspecified problem to the causal target parameter.  In general, the pseudo-true value has an interpretation as a projection.  Standard implementation of generalized empirical likelihood might fail to recover the parameter of interest, and an adjusted procedure is analyzed.  The limiting distribution of the estimator is consistent but potentially non-Gaussian.

 

Location:

Harrison 170