Statistical and Methodological Red Flags in 'Study on Port Logistics Efficiency and Its Influencing Factors in the Beijing-Tianjin-Hebei Region' (Zhang et al., Jiangsu Shanglun, 2020)

Academic fraud report · Geng Detector

Summary

This report documents severe statistical and presentational problems in a 2020 paper published in Jiangsu Shanglun (DOI: 10.13395/j.cnki.issn.1009-0061.2020.07.007) by Zhang Yang and colleagues from Tianjin University of Commerce. The verdict is 'confirmed serious issues' (实锤), though the assessment is constrained by the fact that only the text extract was available for analysis. The most critical finding is a fatal overfitting problem: the authors fit seven independent variables plus a constant to a time series with only n = 11 annual observations (2007–2017), leaving just 3 residual degrees of freedom. SPSS output in Table 6 contains numerous '0.0' standard errors and near-zero coefficients that produce finite t-values, an arithmetic impossibility indicating unformatted copy-paste of software output. The authors claim Min-Max normalization yet report coefficients on the order of 1e-5 to 1e-8, incompatible with [0,1] inputs. Average values in Table 5 also show inconsistent rounding. Image-level forgery could not be assessed due to missing figures.

Verdict

🔴 Confirmed serious issues — The paper contains multiple statistical and methodological errors that render the regression analysis effectively meaningless. Severity ranges from yellow (data arithmetic) to red (overfitting, mathematically contradictory output). Image-level forensics were not possible owing to missing figure files.

Key findings

Fatal overfitting (Finding 1): A multivariate regression with 7 independent variables and an intercept is fit to n = 11 annual observations (2007–2017), leaving residual df = 3. Significance values are statistically unreliable by construction.
Schrödinger's significance (Finding 2): Table 6 displays standard errors of 0.0 for several variables while simultaneously reporting finite t-values (e.g., 4.272, -4.43, -5.516). Mathematically, 0/0 is undefined, so any finite t-value contradicts a true zero standard error; the figures almost certainly reflect rounded outputs from SPSS that the authors pasted without reformatting (e.g., true values near 0.004 or coefficients near -0.000043).
Normalization claimed but not applied (Finding 3): The text states that Min-Max normalization onto [0,1] was performed, but unstandardized coefficients of 3.266E-5, 6.491E-6, and 2.959E-8 are inconsistent with [0,1] regressors producing effects on a mean ~1.0 response. Either the data were not normalized as described, or the model specification is erroneous.
Inconsistent averaging in Table 5 (Finding 4): Recalculation of the 2015 row — (4.95 + 0.00 + 1.26 + 1.25)/4 = 1.865 — is reported as 1.86, whereas 2010 (sum 0.8475) is rounded up to 0.85 and 2013 (sum 0.8575) is rounded up to 0.86. The rounding convention is inconsistent across rows.
Image analysis limited (Finding 5): No high-resolution figures were available, so pixel-level, splicing, or axis-truncation checks could not be performed.

Evidence highlights

DOI: 10.13395/j.cnki.issn.1009-0061.2020.07.007
Journal: 江苏商论 (Jiangsu Shanglun), received 2020-05-03, published July 2020.
Sample: 2007–2017 annual macro data for the Beijing-Tianjin-Hebei region → n = 11.
Regression: 7 IVs + intercept → df_res = 11 − 7 − 1 = 3.
Table 6 examples (as printed): 政府扶持力度 B = 0.0, SE = 0.0, t = -5.516, p = 0.012; 经济发展水平 SE = 0.0; 环境治理投资 SE = 0.0; 集疏运能力 SE = 0.0.
Table 5, 2015 means: Tianjin 4.95, Qinhuangdao 0.00, Tangshan 1.26, Huanghua 1.25; computed average = 1.865, reported = 1.86.
Table 5, 2010 means: 1.30, 0.24, 1.83, 0.02; computed = 0.8475, reported = 0.85 (rounded up).
Table 5, 2013 means: 1.29, 0.03, 0.32, 1.79; computed = 0.8575, reported = 0.86 (rounded up).

Notes

All findings are derived from the textual extract of the PDF; the authors' underlying SPSS .sav file and Output were not examined and should be requested for verification.
Reported p-values are taken at face value from Table 6; no statistical software was re-run.
The report characterizes presentational and methodological defects, not deliberate intent. A formal misconduct determination requires institutional investigation.
The strongly worded commentary attributed to 'Geng' in the source material is the original Chinese author's editorial voice and has been paraphrased here only to keep the technical findings central.