Key points are not available for this paper at this time.
Abstract Existing research recognises that the relationship between global life satisfaction (GLS) and domain-specific life satisfaction (DLS) is more complex than those explained by traditional linear models. This study explores the complexity of the GLS-DLS relationship in early adolescents using various statistical models and investigates the relative importance of key life domains contributing to adolescent well-being. The study used the fourth-wave cross-sectional survey data from the International Survey of Children’s Well-Being, which comprises 22,200 observations from early adolescents (51% girls) aged 7–15 years from 20 countries or territories. The relationship between GLS and satisfaction with 12 key life domains was explored using three traditional regression models and one machine learning approach. A series of goodness-of-fit indicators were used to identify the best model. The machine learning approach provided higher explanatory power in understanding the GLS-DLS relationship than the counterpart models using the full sample, and the linear additive model exhibited better accuracy when predicting the GLS in the cross-validation. Among 12 life domains, health, future, possessions, time use, and safety are the top five important domains for early adolescents. However, heterogeneities in the relative importance of the key domains were observed across countries at different economic development levels. Meanwhile, the advantage of a more advanced machine learning approach provides richer information on explaining the complex relationship within the sample. Moreover, policies aimed at improving the well-being of early adolescents through domain-specific life satisfaction should be tailored to each country’s unique context.
Winn et al. (Mon,) studied this question.