MULTIVARIATE STATISTICAL TECHNIQUES FOR ANALYZING SOCIOECONOMIC AND DEMOGRAPHIC DATA PATTERNS

Abstract

Socioeconomic inequality is increasingly recognized as a multidimensional phenomenon shaped by the interaction of economic, demographic, and household-related factors rather than a single linear hierarchy. Traditional univariate and index-based approaches often fail to capture this structural complexity, leading to oversimplified representations of social stratification. To address this limitation, this paper adopts an integrated multivariate analytical framework combining principal component analysis (PCA), unsupervised clustering, nonlinear Isomap embedding, and profile-based visualization. Using cross-sectional socioeconomic and demographic data, the analysis reveals that variation is distributed across multiple latent dimensions rather than dominated by a single axis of advantage. The clustering results indicate the presence of typical socioeconomic profiles rather than sharply bounded groups, a pattern consistent with continuous rather than categorical social differentiation. Nonlinear embedding further supports this interpretation by highlighting gradual transitions between profiles. Profile visualizations translate these abstract patterns into substantively interpretable configurations, illustrating how distinct forms of advantage and constraint coexist. Overall, the findings demonstrate that socioeconomic positioning is relational, multidimensional, and context-dependent. Methodologically, the study provides a robust alternative to single-index models and offers a scalable template for future research on complex social inequalities.