Answers are grounded in this paper and cite their sources.
Theoretical analysis demonstrates asymptotic equivalence between plug-in and known-boundary estimators in fixed-breakpoint models, indicating robust large-sample inference.
The two-step estimation procedure in structural break detection gives rise to a post-selection inference problem: when the regime boundary is unknown and is first estimated by a test statistic, and the switching degree is then computed on either side of the estimated boundary, the plug-in estimator Ŝ_plug differs from the estimator Ŝ_known obtained with a known boundary. In this paper we prove, in the fixed-breakpoint model (where the switching difference vector Δθ is fixed and nonzero), a core theoretical result: the asymptotic equivalence theorem. Under assumptions A1–A7, √n (Ŝ_plug − Ŝ_known) = o_p(1), and hence √n (Ŝ_plug − S) →_d N(0, V₀), exactly as in the known-boundary case. The source of this result is the n-consistency of the boundary estimator in the fixed-breakpoint model: the number of misclassified samples is O_p(1), their effect on the estimated coefficients is O_p(n⁻¹), and this is invisible at the √n scale. Building on this result, we discuss finite-sample correction directions: (i) a first-order heuristic expansion of the conditional bias, indicating that the post-selection bias, although invisible at the √n scale, may be relevant in finite samples; (ii) a fixed-design residual bootstrap scheme, proposed as a working hypothesis, which keeps the design points (Tᵢ, Xᵢ) fixed and resamples only residuals, aiming to mimic the uncertainty of the boundary selection process; (iii) asymmetric bootstrap quantile confidence intervals as a construction proposal. We strictly distinguish proved theorems from unproved working hypotheses. The paper is purely theoretical; numerical validation is an open problem. Research Paradigm Statement: The core methodology, research direction, and final decisions were independently directed by the author. DeepSeek assisted with code implementation, data presentation, and text drafting. The author takes full academic responsibility for the final content.
No takes yet. Share an insight, caveat, or question.
Shuiping Tang (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: