Asymptotic properties of Pocock–Simon minimization

Characterize the asymptotic properties of Pocock–Simon minimization, particularly the asymptotic behavior relevant to covariate-adjusted distribution and survival-function estimators under this covariate-adaptive randomization scheme.

Background

The paper establishes that the proposed empirical-likelihood estimators have asymptotic distributions invariant to the randomization scheme when the randomization procedure satisfies condition (D) and the randomization covariate is included among the adjustment covariates. The authors explicitly identify Pocock–Simon minimization as an important exception in the literature because its asymptotic behavior is not yet adequately understood.

This unresolved issue matters because Pocock–Simon minimization is a commonly used covariate-adaptive randomization procedure. Establishing its asymptotic properties would clarify the theoretical basis for inference and variance estimation under this design and determine the scope of the claimed randomization-invariance results.

References

Theorem 1 shows not only the asymptotic validity of covariate adjusted $ F_j$ and $ S_j$ but also the invariance of their asymptotic distributions, i.e., the same formula holds for simple randomization or any covariate-adaptive randomization satisfying (D) including Pocock-Simon's minimization whose asymptotic property is still not well understood.

— Shape-Preserving Covariate Adjustment via Empirical Likelihood in Randomized Experiment  (2608.19423 - Lou et al., 19 Aug 2026) in Section 4, elaboration following Theorem 1, item 1

The source of the variance inflation originates because the assignment probability depends only on $\operatorname{sgn}(\Lambda\psi(\vZ))$. However, the updated covariate imbalance, $\Lambda$, is determined by the magnitude of $\psi(\vZ)$. We therefore conjecture that a smooth assignment function may mitigate this problem, but it is left for future research.

— Discretization in covariate-adaptive randomization: gains and losses  (2609.11012 - Zhao et al., 10 Sep 2026) in Section 2, immediately following Proposition 2.1 (the proposition labeled “Less is more”)