Addressing measurement issues in affect dynamic research: Modeling emotional inertia’s reliability to improve its predictive validity of depressive symptoms.

Methodical developments facilitated research on the time-dynamic nature of emotions, introducing novel emotion dynamic measures such as emotional inertia that initially showed significant associations with well-being outcomes like depressive symptoms. However, recent research has challenged this notion by demonstrating that negative emotion inertia's explanatory power in predicting depressive symptoms vanished once mean negative emotion was controlled for. Emotional inertia is often modeled by a two-step approach that first derives estimates of emotional inertia and then uses those to predict depressive symptoms. In the present research, we reanalyzed five experience sampling data sets (N = 875 participants) and demonstrate that this two-step approach leads to low reliability of negative emotion inertia, r¯sb = .52; thereby, attenuating its association with depressive symptoms, as reflected by only 1.3% added explained variance in depressive symptoms above mean negative emotion. As an alternative, we propose a novel one-step approach that adjusts for unreliability of inertia estimates: We introduce a latent inertia factor that is defined by the autocorrelation of various emotion items. Using dynamic structural equation models, this latent factor is simultaneously used to predict depressive symptoms. Here, negative emotion inertia showed good reliability, ω¯ = .81, and explained an additional 4.5% of the total variance in depressive symptoms. Thus, our results demonstrate that emotion dynamic measures can be an important feature of individual well-being if their lower reliability compared with mean negative emotion is modeled and corrected for in dynamic structural equation models. (PsycInfo Database Record (c) 2023 APA, all rights reserved).