Structural equation modeling of adolescents' life satisfaction

Objectives: The aim of the study was to test the effect of emotional support and family cohesion on the relationship among life satisfaction, somatic/psychic problems and adaptation to school environment. Subjects and setting: The subjects consisted of 921 15-year-old Czech pupils (508 girls and 413 boys). The data, extracted from the Health Behaviour in School-aged Children International data archive, were collected using structured questionnaire in school environment. Statistical analysis: Relations among variables based on an estimate of the asymptotic sampling covariance matrix of the polychoric sample correlations, were analyzed by structural equation modeling using Generally Weighted Least Square iterative procedure. Results: Analysis of data using SEM brought acceptable support for the proposed model assuming effect of emotional support and family cohesion on the relationships among life satisfaction, somatic/psychic problems and adaptation to school environment. Model accounted for 31% and 35% of variance of boys and girls life satisfaction, respectively. The magnitude of congruence between the model and the data was noteworthy. Indices reflect that the model fits well and was also, apart from interfactor correlations and notably from error variances, gender invariant. Standardized effects of emotional Support and family cohesion on the life satisfaction, and family cohesion on adaptation to school environment were very similar and statistically significant for both groups. Nevertheless, it seems that life satisfaction of girls is noticeably more strongly associated with the degree of perceived emotional support, and also their adaptation to school environment could be more potent source of somatic/ psychic problem in comparison to boys. Study limitation: Structural equation models are essentially correlational in nature. Consequently, even with adequate fit indices and the explanation of a large proportion of variance in the criterion construct, other models could be equally viable. The present study is subject to several limitations. First, the findings require replication with an independent sample. Second, because the data relied on self-report measures, replication studies that include third-party observations of the participants' behavior would be desirable.