Fig. 1

Dynamic structural equation modeling of arousal and self-esteem. Symbols in bold indicate effects relevant for our tests of hypotheses. Arousal and self-esteem are decomposed into their respective between-person (\({\mu }_{A{rousal}_{i}}\) and \({\mu }_{{Self-esteem}_{i}}\)) and within-person (\({\mu }_{A{rousal}_{it}}\) and \({\mu }_{{Self-esteem}_{it}}\)) parts. The between-person parts reflect the daily average scores for arousal and self-esteem across all time points t for participant i. On the within-level, the model includes random autoregressive effects (φ_iArousal and φ_{iSelf−esteem}) and random cross-lagged effects (ϕiAonS and ϕiSonA) of order 1 on the within-person level. The autoregressive effects reflect the average daily regression coefficient of arousal and arousal at the next time point (φ_iArousal) or self-esteem and self-esteem at the next time point (φ_{iSelf−esteem}) for participant i. The cross-lagged effects reflect the average daily regression coefficient of self-esteem and arousal at the next time point (φ_{iSelf−esteem→Arousal}) or arousal and self-esteem at the next time point (φ_{iArousal→Self−esteem}) for participant i. In addition, we modeled the variance of the innovation terms of arousal (\({\zeta }_{A{rousal}_{it}}\)) and self-esteem (\({\zeta }_{{Self-esteem}_{it}}\)) as random effects. Innovation variances depict unexplained variance due to measurement error and interindividual differences in the reactivity to unobserved influences