A dual-system, machine-learning approach reveals how daily pubertal hormones relate to psychological well-being in everyday life

• Published in: Developmental cognitive neuroscience 2022-12, Vol.58, p.101158, Article 101158
• Main Authors: Chafkin, Julia E., O’Brien, Joseph M., Medrano, Fortunato N., Lee, Hae Yeon, Josephs, Robert A., Yeager, David S.
• Format: Article
• Language: English
• Subjects: Adolescence; Adolescent; Bayes Theorem; Development; Endocrinology; Hormones; Humans; Machine Learning; Mental Health; Original Research; Psychological Well-Being; Psychopathology; Stress, Psychological - psychology
• ISSN: 1878-9293; 1878-9307; 1878-9307
• DOI: 10.1016/j.dcn.2022.101158

The two studies presented in this paper seek to resolve mixed findings in research linking activity of pubertal hormones to daily adolescent outcomes. In study 1 we used a series of Confirmatory Factor Analyses to compare the fit of one and two-factor models of seven steroid hormones (n = 994 participants, 8084 samples) of the HPA and HPG axes, using data from a field study (https://www.icpsr.umich.edu/web/ICPSR/studies/38180) collected over ten consecutive weekdays in a representative sample of teens starting high school. In study 2, we fit a Bayesian model to our large dataset to explore how hormone activity was related to outcomes that have been demonstrated to be linked to mental health and wellbeing (self-reports of daily affect and stress coping). Results reveal, first that a two-factor solution of adolescent hormones showed good fit to our data, and second, that HPG activity, rather than the more often examined HPA activity, was associated with improved daily affect ratios and stress coping. These findings suggest that field research, when it is combined with powerful statistical techniques, may help to improve our understanding of the relationship between adolescent hormones and daily measures of well-being.