Curriculum Vitae

This paper proves ratio-consistency of the jackknife variance estimator for a broad class of Hájek-dominated generalized U-statistics, including the classical fixed-order case. It also delivers weaker sufficient conditions for consistent variance estimation in two-scale distributional nearest-neighbor regression.

This paper develops a localized distributional nearest-neighbor framework for pointwise inference on conditional moment targets, including heterogeneous treatment effects and conditional quantile slopes. It combines orthogonal scores with leave-one-out stability in place of cross-fitting and yields feasible inference using a jackknife variance estimator and a localized finite-difference Jacobian.