The paper “Two-Step Semiparametric Empirical Likelihood Inference”, coauthored with F. Bravo and I. Van Keilegom, and published as leading article in The Annals of Statistics, 48(1), 1-26, 2020, studies inference for structural parameters in the presence of high-dimensional first steps. A leading example is inference on the average treatment effect of a policy when the probability of participation in the policy is estimated in a first step. Another example is inference on structural parameters when some data is missing, and nonparametric imputation methods are used for the missing data. Previous research based on empirical likelihood in this setting relied on bootstrap approximations and complicated asymptotic distributions. We propose a methodology that leads to simple asymptotic distributions (chi-square) without the need to use bootstrap.
The paper is available online: https://projecteuclid.org/euclid.aos/1581930123