本研究基於Kaggle平台提供的2015年BRFSS資料庫，針對不同的身體健康狀況與年齡層劃分子樣本異質性，並設計了一套基於熵模擬的探索性類別資料分析方法，旨在深入探討慢性疾病(心臟病與中風)與其他風險類別之間的關聯性。研究透過列聯表，分別在虛無假設與對立假設下進行熵值模擬，計算型一錯誤與型二錯誤的最小總和，並結合可靠性檢查與演算法，找出 24 組子樣本中1-階與2-階風險類別的交互作用。
我們對1-階疾病主風險類別進行階層分群，並以熱圖方式展示，從不同身體健康狀況與年齡層的雙重視角進行解讀。同時，我們以路線圖呈現二階疾病主風險類別的動態變化，並進一步區分為正向與負向類別進行探討。研究結果顯示，2-階交互效應可分為三種類型：對稱型、方向型與正負切換型，其中負向疾病風險類別的演化軌跡普遍比正向疾病風險類別更為複雜。本研究不僅揭示了疾病風險的動態演變特徵，還提供了一種有效的分析框架，為慢性疾病的預防與風險管理提供了新的視角與數據支持。

This study, based on the 2015 BRFSS dataset provided by the Kaggle platform, focuses on segmenting sample heterogeneity by different health conditions and age groups. It introduces an exploratory categorical data analysis method based on entropy simulation to deeply investigate the relationships between chronic diseases (heart disease and stroke) and other feature categories. Using contingency tables, entropy simulations were conducted under null and alternative hypotheses to calculate the minimal total of Type I and Type II errors. This approach, combined with reliability checks and algorithms, identified the interactions of major 1-featurre and major 2-feature categories across 24 subgroups. We performed hierarchical clustering of first-order major risk categories and visualized them using heatmaps, providing interpretations from the dual perspectives of general health conditions and age groups. Additionally, we used roadmaps to depict the dynamic changes of major 2-feature categories, further distinguishing them into positive and negative categories. The study revealed that second-order interaction effects can be classified into three types: symmetric, directional, and positive-negative switching. Among these, the evolutionary trajectories of negative disease risk categories were generally more complex than those of positive disease risk categories. This research not only highlights the dynamic evolution characteristics of disease risks but also provides an effective analytical framework, offering new perspectives and data support for the prevention and management of chronic diseases.

高宏維（2024）。以基於熵值模擬之演算法探討類別資料中的複雜關聯（碩士論文，國立政治大學）。國立政治大學統計學研究所。
Shannon, C. E. (1948). A mathematical theory of communication. Bell System Technical Journal, 27(3), 379–423.
Wilkinson, L., & Friendly, M. (2009). The history of the cluster heat map. The American Statistician, 63(2), 179–184.
Fushing, H., Chou, E. P., & Chen, T.-L. (2023). Multiscale major factor selections for complex system data with structural dependency and heterogeneity. Physica A: Statistical Mechanics and its Applications, 630, 129227.
Fushing, H., Kao, H.-W., & Chou, E. P.-T. (2024). Topological risk-landscape in metric-free categorical database. IEEE Access. https://doi.org/10.1109/ACCESS.2024.3398416
Centers for Disease Control and Prevention. (2014). About BRFSS. Centers for Disease Control and Prevention.
Kolmogorov, A. N. (1983). On logical foundations of probability theory. Lecture Notes in Mathematics. https://www.cdc.gov/brfss/about/index.htm
Chen, T.-L., Chou, E. P., & Hsieh, F. (2022). Categorical nature of major factor selection via information theoretic measurements. Entropy, 23(12), 1684.
Chen, C., & Fushing, H. (2012). Multi-scale community geometry in network and its application. Physical Review E, 86, 041120.
Fushing, H., & Chen, C. (2014). Data mechanics and coupling geometry on binary bipartite network. PLoS ONE, 9(8), e106154.