AUTHOR(S): Adebayo O. P. Ahmed. I, Oyeleke K. T.
|
TITLE |
![]() |
ABSTRACT This study examines the complex relationship between political ideology and religious values in the United States through a comprehensive machine learning framework. Analyzing data from 9,349 respondents across six waves of the World Values Survey (1982-2011), we compared multiple classification approaches for predicting political ideology while investigating key determinants of religious importance. Our findings reveal that ensemble methods substantially outperformed specialized ordinal techniques, with Random Forest achieving 32.3% accuracy in ideology prediction compared to ordinal regression's 9.7% performance. The LASSO regression analysis demonstrated remarkable variable selection parsimony, identifying only core religious importance as a meaningful predictor while eliminating all demographic and attitudinal variables from the final model. Principal Component Analysis revealed multidimensional structure in social attitudes, with no single dominant dimension emerging. Temporally, political ideology maintained remarkable stability across three decades, fluctuating within a narrow 5.72-5.93 range, while religious importance showed a gradual decline from 8.31 to 7.76. These results challenge conventional methodological assumptions about ordinal data analysis and provide new insights into the evolving landscape of American political and religious attitudes, suggesting that practical predictive accuracy may outweigh theoretical model specifications in complex social measurement contexts. |
KEYWORDS Political Ideology Prediction, Ordinal Classification, Religious Importance, Machine Learning Comparison, American Public Opinion, World Values |
|
Cite this paper Adebayo O. P. Ahmed. I, Oyeleke K. T.. (2025) Modelling Political Ideology and Religious Importance in the US: An Ordinal and Binary Classification Analysis. International Journal of Biology and Biomedicine, 10, 58-72 |
|