Scholarlink

<p>This paper examines the determinants of subjective well‑being using a mixed quantitative approach that combines classical econometric modelling with machine‑learning and dimensionality‑reduction techniques. Drawing on data from a large nationally representative survey conducted in Spain, we estimate an ordinary least squares model to identify average associations between life satisfaction and a set of socioeconomic, demographic and attitudinal variables. To complement this analysis, a Random Forest model is employed as a non‑parametric validation tool to assess the robustness of the relative importance of predictors beyond linear assumptions. In addition, principal component analysis and k‑means clustering are used to construct data‑driven social profiles that capture heterogeneity in well‑being outcomes. Across methods, indicators of economic vulnerability and subjective social position emerge as the most consistent correlates of life satisfaction, while several demographic characteristics display weaker or context‑dependent associations. The comparison between parametric and non‑parametric results illustrates how different quantitative strategies can converge on similar substantive conclusions while offering complementary insights into complex social phenomena. By integrating econometric estimation, machine learning and clustering techniques within a single analytical framework, the study contributes to the methodological literature on the quantitative analysis of quality‑of‑life data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A mixed econometric and machine‑learning approach to the analysis of subjective well‑being: evidence from a national survey

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月31日

查看全文
  • Antonio Jurado,
  • Marcelo Sánchez-Oro Sánchez,
  • Rafael Robina-Ramírez

摘要

This paper examines the determinants of subjective well‑being using a mixed quantitative approach that combines classical econometric modelling with machine‑learning and dimensionality‑reduction techniques. Drawing on data from a large nationally representative survey conducted in Spain, we estimate an ordinary least squares model to identify average associations between life satisfaction and a set of socioeconomic, demographic and attitudinal variables. To complement this analysis, a Random Forest model is employed as a non‑parametric validation tool to assess the robustness of the relative importance of predictors beyond linear assumptions. In addition, principal component analysis and k‑means clustering are used to construct data‑driven social profiles that capture heterogeneity in well‑being outcomes. Across methods, indicators of economic vulnerability and subjective social position emerge as the most consistent correlates of life satisfaction, while several demographic characteristics display weaker or context‑dependent associations. The comparison between parametric and non‑parametric results illustrates how different quantitative strategies can converge on similar substantive conclusions while offering complementary insights into complex social phenomena. By integrating econometric estimation, machine learning and clustering techniques within a single analytical framework, the study contributes to the methodological literature on the quantitative analysis of quality‑of‑life data.