Integrity Check Report: The Bayes Factor and Its Implementation in JASP: A Practical Primer (DOI: 10.3724/SP.J.1042.2018.00951)

Academic fraud report · Geng Detector

Summary

This AI-assisted integrity review of Hu Chuan-Peng et al. (2018), 'The Bayes Factor and Its Implementation in JASP: A Practical Primer', published in Advances in Psychological Science (DOI: 10.3724/SP.J.1042.2018.00951), concludes the paper is CLEAN, with no evidence of academic misconduct. The only confirmed defect is a minor figure-reference typo in Section 3.2 ('Figure 1' should read 'Figure 2'), an editorial slip, not fraud. Numerical values (BF01 = 3.71, BF01 = 7.74, BF0+ = 7.73, BF0+ = 10.75, r = 0.39, p = 0.007, 95% CI [-0.5, 0.23]) were verified as internally consistent and traceable to public data. Forensic image signals (copy-move at offset (32,0), noise variance inconsistency, shared PRNU) are explained by JASP software screenshots explicitly described in captions, not experimental imagery. A statistical column-difference flag was deemed inapplicable to qualitative comparison tables. Confidence is moderate-to-high given the methodological nature of the paper; final determinations require institutional review.

Verdict

✅ CLEAN — No evidence of academic misconduct. The sole confirmed issue is a figure-number citation typo in Section 3.2 (editorial error). Recommend only a corrigendum to the editorial office; no further investigation warranted.

Key findings

Finding 1 — Figure citation typo (minor, not misconduct): Section 3.2 states “将显示如图 1中间图所示的对话框” but context and subsequent references (“如图2中间所示…γ = 0.707”) indicate Figure 2 was intended. Programmatic check: 2 cited data points, 2 matched in source text. Severity: 🟡 editorial slip only.
Finding 2 — No data fabrication: All reported values (BF01 = 3.71, BF01 = 7.74, BF0+ = 7.73, BF0+ = 10.75, r = 0.39, p = 0.007, median -0.13, 95% CI [-0.5, 0.23]) derive from public demonstration data (Wagenmakers et al., 2015 replication of Topolinski & Sparenberg, 2012) and are internally consistent (BF01 = 3.71 is the reciprocal of BF10 = 0.27, as stated in the paper).
Finding 3 — Image forensics benign: Copy-move matches at fixed offset (32,0), noise-variance inconsistency, and cross-image PRNU commonality are fully explained by embedded JASP interface screenshots (Figures 2–4), whose captions explicitly identify them as software operation/output screenshots. The paper contains no Western blots, gels, microscopy, or experimental imagery.
Finding 4 — No statistical anomalies: Cited statistics come from published sources (Stulp et al., 2013: r = 0.39, p = 0.007; Wagenmakers et al., 2015); the paper itself explicitly criticizes p-hacking and selective reporting (Section 4).
Finding 5 — Timeline consistent: Received 2017-10-10; published June 2018; chinaXiv preprint 201709.00120v4 (September 2017) consistent with publication workflow. No anomaly.
Finding 6 — Citations and methods accurate: References complete (Jeffreys 1961; Rouder et al. 2009; Wagenmakers et al. 2017); default prior γ = 1/√2 ≈ 0.707 stated correctly; JASP Version 0.8.2 (2017) predates submission.
Forensic flag [S1] inapplicable: “Minimal column-difference variance” flag arises from qualitative comparison tables (√/× symbols, BF intervals) and reference-number columns, not quantitative data columns; test precondition not met.

Evidence highlights

Section 3.2: “将显示如图 1中间图所示的对话框” vs. later “如图2中间所示，将Cauchy先验的宽度设置为 JASP的默认值γ = 0.707” — typo confirmed.
Figure 2 caption explicitly: “使用JASP进行贝叶斯独立样本 t检验时的操作截屏” — screenshots, not experimental data.
Verified numeric matches: BF01 = 3.71 ↔ BF10 = 0.27; one-sided BF01 3.71 → 7.74; robustness BF0+ = 7.73 (r = 0.707) and BF0+ = 10.75 (r = 1).
Programmatic evidence verification: 2/2 cited data points matched.

Notes

This report is AI-assisted and intended for academic discussion only. Final determination of misconduct requires investigation by the relevant institutions;误报/漏报 possible. Authors' reputational rights are respected. If disputed, official investigation conclusions prevail.
Suggested action: notify the editorial office of the figure-reference typo (potential erratum); no other action needed.