Extended Data Fig. 6: Structural equation model, longitudinal analysis and mediation analysis for participants with more than 7 hours sleep.
a. The longitudinal association between the sleep duration, depression and fluid intelligence revealed by cross-lagged panel model. The baseline sleep duration (β= 0.025, p = 1.3 × 10−5) and depressive symptom (β= −0.023, p = 1.3 × 10−5) was significantly associated with fluid intelligence in the follow-up. b. Mediation analysis. The mediation models were conducted to analyze the direct relationship between sleep duration and fluid intelligence, with sleep duration, brain structure and both of them as mediator respectively. Brain regions significantly mediated the association between sleep duration and fluid intelligence (β= −0.0046, p = 1.4 × 10−5). These figures utilized participants with sleep duration more than 7 hours. c. Full frame model. Standardized coefficients were showed in the figure. PRS was significantly associated with mental health (β = −0.034, p = 4.7 × 10−5). Brain volumes were a better predictor of cognitive function (β= −0.198, p < 1.0 × 10−20) compared to mental health (β = 0.048, p = 3.5 × 10−6). Sleep duration was the most significant predictor of mental health (β = 0.167, p < 1.0 × 10−20) and brain regions (β= −0.044, p < 1.0 × 10−20). Latent variable including brain structure, mental health and cognitive function were estimated in the model which showed in the figure with orange, green and blue box respectively. Wald tests were utilized to derive the two-sided p value adjusted for multiple comparisons (FDR correction). * represented p < 0.05, ** represented p < 0.01 and *** represented p < 0.001.
