File:The longitudinal association and mediation analysis on the relations between early 'reading for pleasure' and youth cognitive and psychopathological symptoms scores.png

English: "(a, b) The structural equation analysis using the CLPM model indicated a longitudinal association of the early RfP with cognitive as well as attention problems scores. Higher levels of early RfP recorded at baseline were associated with better cognition crystallized scores (β = 0.238, p < 1.0 × 10–4, s.e. = 0.024) and lower attention symptoms (β = −0.031, p = 0.01, s.e. = 0.012) 2-years later, and the reverse directions were also significant. (c–f) Mediation analysis findings: the indirect path (A, AB, and B) indicated that the brain cortical area significantly mediated the associations between early RfP and young adolescent cognitive (c, d) and psychopathological scores (e, f). (bias-corrected p and bootstrap p < 0.001). Path_A indicates the direct effect of the predictor factor (independent variable, early RfP) with the mediator (youth brain cortical area structure: the mean cortical area of the significant brain areas shown in Fig. 2f or Fig. 2i); Path_B indicates the direct effect of the brain structural mediator with the dependent variable (youth clinical assessments: cognitive or psychopathological scores); Path_C indicates the total effect between the predictor factor and the clinical assessments when the mediator was taken into account, meanwhile Path_C’ indicates the direct effect, controlling for the mediator. Path_AB is the product of path_A and path_B (βpath_AB = βpath_A × βpath_B), indicating the mediation effect between the predictor factor and the assessments outcomes through the brain cortical structures. The mediation effects of path_AB implemented by increased cortical area (red colour) between early RfP and youth higher cognitive (orange colour) and lower psychopathological scores (blue colour) were all significant. The β values represent regression coefficients of the effect of the independent variables on the dependent variables. Statistical tests were two-sided. pBonferroni < 0.05. CLPM, two-wave cross-lagged panel model; s.e., represents standard error."