Chapter 5 156 Additional (Non-Preregistered) Analyses We additionally explored whether the proportions of families showing reciprocal, parentdriven, adolescent-driven, and null effects were different from what would be expected by chance (i.e., 25% of families showing each type of effect). To do so, we used chisquare tests, reviewing standardized residuals (≥ |1.96|) to interpret which effects were significantly more or less prominent in our sample. Moreover, we tested whether demographic factors and the two personality traits could explain differences in terms of (absolute) effect sizes. Specifically, we tested for sex differences (t-test), differences between adolescents with varying educational levels (ANOVA), and correlations with age and trait levels of environmental sensitivity and neuroticism. Figure 4 Specification of Dynamic Structural Equation Model P t-1 P t A t-1 A t ϕPP ζA,t ζP,t ϕAP ϕPA ϕAA Decomposition Within Between P t A t P t A t μP μA * * * * * * ϕPP ϕAP ϕPA ϕAA μP μA ML-VAR(1) Pre-registered Note. P = Parenting. A = Adolescent affect. Left: Variables are decomposed in a between-family (μ = family-specific mean) and within-family part (P*t and A*t = time-specific score of parenting and adolescent affect, respectively). Top right: Estimates at the within-family level, including the random (family-specific) cross-lagged (ϕAP and ϕPA) and autoregressive effects (ϕPP and AA) and the correlation between the innovations (ζ). Bottom right: Between-family level correlations between the random effects and means.
RkJQdWJsaXNoZXIy MTk4NDMw