Per-draw robust linking

Research Method · All research methods
Evidence · 1 claim (1 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 1 claim rests on one study

Description

The linking step maps each independently calibrated period to a common reference metric by minimizing a response-count-weighted pseudo-Huber (robust Haebara) discrepancy between expected-score curves on anchor items. Following Baldwin (2011), the criterion is minimized separately for each posterior draw rather than once at the posterior mean; the article states that "Applying it draw-wise propagates the linking uncertainty, together with its dependence on the item parameters, into the linked posteriors." Weighting anchors by response count lets better-estimated items contribute more to the link, approximating concurrent calibration.

Accounts

• Per-draw robust linking: solving a response-count-weighted robust Haebara criterion separately for each posterior draw propagates linking uncertainty into the linked posteriors: The linking step maps each independently calibrated period to a common reference metric by minimizing a response-count-weighted pseudo-Huber (robust Haebara) discrepancy between expected-score curves on anchor items. Following Baldwin (2011), the criterion is minimized separately for each posterior draw rather than once at the posterior mean; the article states that "Applying it draw-wise propagates the linking uncertainty, together with its dependence on the item parameters, into the linked posteriors." Weighting anchors by response count lets better-estimated items contribute more to the link, approximating concurrent calibration. (Paul A. Jewsbury et al. (2026))

Claims

• Consensus calibration closely reproduces pooled-benchmark item-parameter means and standard deviations on an operational assessment [+M]

Related Research Methods

-

Key Sources

• Paul A. Jewsbury and Steven W. Nydick and Manqian Liao and Siyuan (Marco) Chen. (2026). Bayesian Consensus Calibration of Continuously Evolving IRT Item Banks. https://arxiv.org/abs/2609.13590v1

Page metadata

Field	Value
type	research-method
id	per-draw-robust-linking
title	Per-draw robust linking
description	The linking step maps each independently calibrated period to a common reference metric by minimizing a response-count-weighted pseudo-Huber (robust Haebara) discrepancy between expected-score curves on anchor items.
status	review
generated	by process:settle-candidates on 2026-10-10
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.