🤖 AI Summary
This study addresses the failure of standard influence function–based inference in finite samples under “near-boundary” settings of semiparametric models, where second-order remainder terms non-negligibly contribute to sampling variance. The authors propose a finite-sample variance decomposition framework that separates influence function variance from remainder-induced variance and establish necessary and sufficient conditions for the consistency of sandwich variance estimators. Building on this framework, they develop two robust variance estimators—the leave-one-unit-out jackknife and a paired-cluster bootstrap—and derive an analytical expression for the interaction between remainder terms and within-cluster correlation in clustered data. Their approach enables valid confidence interval construction near the boundary: the jackknife Wald interval is numerically equivalent to a bias-corrected sandwich estimator and accurately captures the mechanism by which clustering amplifies variance estimation bias.
📝 Abstract
Asymptotically linear estimators in semiparametric models achieve their point-estimation guarantees via a von Mises expansion in which a second-order remainder is declared negligible. Confidence intervals then treat the first-order influence-function term as the sole source of sampling variability. This reasoning is asymptotically exact but can fail materially in finite samples whenever the second-order remainder contributes variation of the same order as the influence-function variance -- a regime we call the \emph{near-boundary regime}, characterized by nuisance estimation operating at or near the product-rate threshold. We develop a general theory of inference for this regime. Our contributions are: (i) a \emph{finite-sample variance decomposition} that separates influence-function variance from remainder-induced variance and the covariance between them; (ii) a \emph{sandwich consistency theorem} that gives a precise necessary and sufficient condition -- strong remainder negligibility -- for the standard sandwich to be consistent for the total sampling variance, and shows this is strictly stronger than the product-rate condition that guarantees asymptotic linearity; (iii) two \emph{refined variance estimators} -- leave-one-unit-out jackknife and pairs cluster bootstrap -- each with full asymptotic validity guarantees in the near-boundary regime, together with a heteroskedasticity-corrected sandwich interpretation that is numerically equivalent to the jackknife Wald interval; and (iv) a \emph{clustered-data extension} in which the remainder interacts with intra-cluster correlation to produce an analytic formula for sandwich gap amplification.