Papers & full reports
Deconstructing the Friday the 13th Effect: Evidence from Dummy-Variable Regression, Wavelet Analysis, and XGBoost Prediction
Scientific Journal of Intelligent Systems Research, 8(5), 42–56. Authors: Leyi Deng, Xiaowei Feng, Yunya Mao, and Zijia Wan. The published paper states equal contributions and shared first authorship.
Published version provided by the journal under CC BY-NC 4.0; authorship, text, and pagination are retained. DOI: 10.54691/rp81mr64.
Deconstructing the Superstition of Friday the 13th: Evidence from Heterogeneous Regression and Machine Learning Feature Engineering
Problem D, team IMMC26559668. The Chinese competition report and the later English publication are separate versions of the same team project.
Turning a belief into a question
Our team examined whether Friday the 13th was associated with negative outcomes in the selected data. We separated the particular date combination from other calendar effects and explored how environmental information could enter a prediction model.
Three parts of the analysis
Dummy-variable regression and comparisons of date effects.
Wavelet analysis to examine variation at different time scales.
XGBoost models using selected features, alongside a composite well-being index.
Team work and publication
I contributed to model development, analysis, writing, and revisions after the competition. The report received an Honorable Mention in the IMMC Greater China Winter Challenge. The later article lists four authors with equal contributions and shared first authorship.
What the comparison can establish
A non-significant result in selected data does not prove that a date can never matter. The interpretation depends on the observations, variables, and models used. A composite well-being index is also a modeling choice, rather than a direct account of every person’s experience.