Paper 4 — predeclared plan v1.0: what makes a model or a prior good or bad for a difference in arithmetic means? (Behrens–Fisher programme)

Prospective registration of the predeclared plan (v1.0, frozen 16 Nov 2026) for Paper 4 in a programme on the direction of small-sample interval failure for a difference in arithmetic means. Paper 4 asks what makes a model or an informative prior good or bad for that difference. On the model side it tests whether the bias of a misspecified family's pseudo-true mean predicts the one-sided error of its posterior interval and its growth with sample size, whether mean-preserving families (gamma, inverse Gaussian) are safe on direction, whether a shape-preserving sandwich calibration repairs coverage, and whether a mean-gap diagnostic guard works. On the prior side, on a registered subset of 18 settings, it tests how shape priors and priors on the difference pull the interval as a function of prior effective sample size and centring error, whether heavy-tailed and robust-mixture priors limit the damage, and what double use of the data does. The design has 156 settings of positive-valued populations new to the programme (13 group pairings, 6 sample-size pairs, 2 mean differences). Hypotheses M1–M4, P1–P4, N1 and R1 are stated with criteria. The package contains the frozen plan (with a typeset PDF copy) and its SHA-256 fingerprint, all code including the frozen confirmatory analysis, the frozen pseudo-true quantities, grid-accuracy and leave-one-out validations, and the Monte Carlo endpoint check that set the draw counts and served as the pilot, all on non-study seeds. No Paper 4 computation on study data had been run at the time of deposit. AI-use note (paste, or use the programme's final wording): The plan, code and checks were developed with Claude (Anthropic) as a computational assistant under the author's direction; the author approved the design decisions and hypotheses and is responsible for the registration.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24279824
Primary Topic
Statistical Methods and Bayesian Inference
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Paper 4 — predeclared plan v1.0: what makes a model or a prior good or bad for a difference in arithmetic means? (Behrens–Fisher programme)

Johan G.L. Verheyden
4 links
Zenodo (CERN European Organization for Nuclear Research)
Statistical Methods and Bayesian Inference
article

Paper 4 — predeclared plan v1.0: what makes a model or a prior good or bad for a difference in arithmetic means? (Behrens–Fisher programme)

Johan G.L. Verheyden
article en

Abstract

Prospective registration of the predeclared plan (v1.0, frozen 16 Nov 2026) for Paper 4 in a programme on the direction of small-sample interval failure for a difference in arithmetic means. Paper 4 asks what makes a model or an informative prior good or bad for that difference. On the model side it tests whether the bias of a misspecified family's pseudo-true mean predicts the one-sided error of its posterior interval and its growth with sample size, whether mean-preserving families (gamma, inverse Gaussian) are safe on direction, whether a shape-preserving sandwich calibration repairs coverage, and whether a mean-gap diagnostic guard works. On the prior side, on a registered subset of 18 settings, it tests how shape priors and priors on the difference pull the interval as a function of prior effective sample size and centring error, whether heavy-tailed and robust-mixture priors limit the damage, and what double use of the data does. The design has 156 settings of positive-valued populations new to the programme (13 group pairings, 6 sample-size pairs, 2 mean differences). Hypotheses M1–M4, P1–P4, N1 and R1 are stated with criteria. The package contains the frozen plan (with a typeset PDF copy) and its SHA-256 fingerprint, all code including the frozen confirmatory analysis, the frozen pseudo-true quantities, grid-accuracy and leave-one-out validations, and the Monte Carlo endpoint check that set the draw counts and served as the pilot, all on non-study seeds. No Paper 4 computation on study data had been run at the time of deposit. AI-use note (paste, or use the programme's final wording): The plan, code and checks were developed with Claude (Anthropic) as a computational assistant under the author's direction; the author approved the design decisions and hypotheses and is responsible for the registration.

Zenodo (CERN European Organization for Nuclear Research)
Statistical Methods and Bayesian Inference
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon