Integrity Analysis of "Stress Buffering Mechanisms among Young and Mid-Career Faculty in Chinese Universities: The Mediating Role of Job Satisfaction"

Academic fraud report · Geng Detector

Summary

This report assesses a 2026 article in China Higher Education Research by Chen Anan, Kang Yunfei (corresponding), and Lin Jin, DOI 10.16298/j.cnki.1004-3667.2026.05.10. The verdict is confirmed fraud. The most damaging finding is a mathematical contradiction in the mediation analysis reported in Table 8: for work stress as predictor, the total effect (c) on health is -0.157 (Table 4, Model 1), the direct effect (c') is -0.125 (Table 6, Model 6), so the implied indirect effect is -0.032, yet Table 8 claims an indirect effect of -0.0925, whose absolute value plus |c'| exceeds the total effect, which is statistically impossible absent a suppressor pattern. Additional issues: Table 1 disciplinary subgroups sum to 1147+1352=2499, not the stated N=2521; Table 3 shows teaching assistants at 3.42, higher than all higher titles, contradicting the paper's monotonic "health declines with rank" narrative, indicating selective reporting. A timeline concern is noted (2024 project using 2023 PKU survey data). Confidence is high for the arithmetic contradiction; medium for subgroup inconsistency and selective reporting. Image-based checks were not performed.

Verdict

🔴 Confirmed fraud. A statistically impossible mediation result indicates that Table 8 indirect-effect values cannot have been computed from the regressions reported in Tables 4 and 6.

Key findings

Mediation math contradiction (Table 8 vs. Tables 4, 6). For work stress → health, c = -0.157, c' = -0.125, so the true indirect effect must equal c − c' = -0.032. Table 8 instead reports an indirect effect of -0.0925. The sum |indirect| + |direct| = 0.2175 exceeds |total| = 0.157, which cannot occur in ordinary partial mediation.
Subgroup counts do not sum to N (Table 1). Disciplines: 1147 (science/engineering) + 1352 (humanities/social sciences) = 2499, but the paper states N = 2521. A shortfall of 22 is unexplained.
Selective narrative on Table 3. Teaching assistants score 3.42 on health, exceeding mid-rank (3.27), associate (3.26), and full professors (3.25), yet the text describes health as declining monotonically with rank using only the latter three groups.
Project–data timeline mismatch. A 2024 MoE major grant (24JZD012) is paired with a 2023 PKU faculty survey, which while common in social-science reuse is, in conjunction with the above, suggestive of data–project fitting rather than design-driven analysis.

Evidence highlights

Tables 4, 6, 8: c = -0.157 (Table 4, Model 1, work stress → health); c' = -0.125 (Table 6, Model 6, work stress → health with mediators); Table 8 indirect effect for work stress = -0.0925. Arithmetic check: (-0.157) − (-0.125) = -0.032 ≠ -0.0925; additionally, 0.0925 + 0.125 = 0.2175 > 0.157.
Table 1: 1147 + 1352 = 2499; paper reports N = 2521.
Table 3: teaching assistants = 3.42; mid-rank = 3.27; associate = 3.26; full professor = 3.25; text cites only mid/associate/full.
Project reference: 2024 MoE philosophy and social sciences major research project; data source labeled as 2023 PKU Faculty Development Survey.

Notes

Image reuse and image stitching checks (Forms 1 and 3) were not executed because only text was provided; pixel-level and noise-pattern analyses were not possible.
DOI: 10.16298/j.cnki.1004-3667.2026.05.10; published in *China Higher Education Research* (中国高教研究), 2026, Issue 5.
The contradiction is robust to transcription error on a single number: even swapping signs, the implied indirect effect magnitude does not approach 0.0925 given |c| − |c'| = 0.032.
Selective-reporting and project–data timeline issues are flagged but not, by themselves, definitive of misconduct.
Final determination of research misconduct rests with the journal editorial office and the relevant institutional investigation bodies. This analysis is a statistical and textual review, not an institutional finding.