Substantive-Methodological Synergy: Novel Methodological Approaches to Long-Standing Issues


​Chair: Joeri Hofmans  (VUB)
This symposium brings together three studies that are characterized by substantive-methodological synergy, or the offering of novel insights through the application of new, advanced methods to substantively important issues. In the first presentation Chiara Carlier will present how the use of a novel measure of profile similarity combined with the application of regression trees allows for a better understanding of predictors of emotional similarity in romantic couples. In the second presentation Tim Vantilborgh applies network models and the idea of bridge symptoms to shed light on the comorbidity of burnout with depression and generalized anxiety disorders. In the final presentation, Jasmine Vergauwe presents a novel theoretical framework and an associated methodology based on bifactor models to study shared and unique perspectives in multisource ratings.

Applying profile similarity and regression trees to chart fluctuations and covariates of emotional similarity in romantic couples
 
Chiara Carlier1, Mirka Henninger², Martine Verhees1, Peter Kuppens1, Eva Ceulemans1
 
1Research group of Quantitative Psychology and Individual Differences, Faculty of Psychology and Educational Sciences, KU Leuven, Belgium
²Center for Statistics & Data Science, Faculty of Psychology, University of Basel, Switzerland
 
Romantic couples are often thought to align their expectations, feel on a same wavelength or to behave ‘in sync’. But when looking at similarity in emotions, we can find only mixed evidence for its existence. Until now, emotional similarity in romantic couples has mostly been examined in a cross-sectional way or by focusing on single variables like positive or negative affect. However, emotional similarity often concerns multiple discrete emotions simultaneously and can be expected to change over time and contexts. We therefore propose to compute the similarity between profiles of multiple discrete emotions on a momentary level, for each timepoint of each couple separately. In this talk, we will use this new framework in two experience sampling studies to examine the fluctuation of momentary emotional similarity. In addition, we will use multilevel regression trees to explore which (combinations of) covariates can best predict this type of similarity and whether being in a romantic relationship makes partners even more similar than what can be predicted with these covariates.
 
Reconceptualizing burnout as a network of symptoms to understand comorbidity with depression and generalized anxiety disorder
 
Tim Vantilborgh1, Valentina, Sagmeister1, Femke Legroux1, Eva Mertens1, Robin Biesmans, Sofie Van Ballaert, Sara De Gieter1
 
1Research group of Work and Organizational Psychology, Faculty of Psychology and Educational Sciences, Vrije Universiteit Brussel, Belgium
 
Burnout, depression, and generalized anxiety disorder (GAD) are significant mental health issues impacting both individuals and organizations, with an estimated annual cost of 940 million euros due to productivity losses. These conditions often co-occur, leading to complex clinical management and poorer health outcomes. Understanding their interconnections can help improve interventions and management strategies. In this study, we applied a network approach to explore the relationships among burnout, depression, and GAD. This method assumes that psychological phenomena arise from causal interactions between symptoms. Specifically, we focused on identifying 'bridging symptoms' that link these conditions, providing insights into why comorbidity occurs. We collected data from 215 Flemish employees using the Burnout Assessment Tool, VierDimensionale KlachtenLijst, Patient Health Questionnaire, and Generalized Anxiety Disorder scale. We generated Gaussian Graphical Models to visualize the interconnections of symptoms across these scales. Our analysis highlighted that 'depression' and 'exhaustion' were central symptoms in our networks, though only depression showed strong connections as hypothesized. Key bridging symptoms included physical characteristics ('slow/restless') linking anxiety and depression; 'mental distance' connecting burnout with depression, but not GAD; and 'difficulty relaxing', 'feeling anxious', 'worrying behavior', 'sleep problems', and 'worrying' serving as links between various conditions. The study confirmed that while burnout, depression, and GAD manifest as distinct conditions, their high comorbidity can be explained by bridging symptoms. This reinforces the utility of a network perspective in understanding and addressing comorbidity. Future research should investigate how these symptoms interact over time to better tailor preventive and therapeutic strategies.
 
The Leadership Arena-Reputation-Identity (LARI) Model: Distinguishing shared and unique perspectives in multisource leadership ratings
Jasmine Vergauwe1, Joeri Hofmans1, and Bart Wille2
 
1Research group of Work and Organizational Psychology, Faculty of Psychology and Educational Sciences, Vrije Universiteit Brussel, Belgium
2Department of developmental, personality and social psychology, Faculty of Psychology and Educational Sciences, Ghent University, Ghent, Belgium
 
Multisource leadership ratings rely on the core assumption that –in addition to the leader’s self-ratings– different rater groups (subordinates, peers and superiors) offer unique perspectives, and thus provide a more well-rounded analysis of the leader’s behavior (i.e., the ‘discrepancy hypothesis’; Borman, 1997). However, current methods provide limited insight into the extent of overlap and uniqueness in these perspectives, leaving the empirical support for the discrepancy hypothesis unclear. In pursuit of improved substantive-methodological alignment, we introduce the Leadership Arena-Reputation-Identity (LARI) model, which conceptualizes both the shared and unique perspectives in terms of latent factors reflecting respectively (i) the consensus about the leader (i.e., the Leadership Arena), (ii) the impressions conveyed to others that are distinct from self-perceptions (i.e., the leader’s Reputation), and (iii) the unique self-perceptions of the leader (i.e., the leader’s Identity). The LARI model is formalized by means of bifactor modeling, which allows to statistically decompose the variance captured by multisource ratings. The LARI model was tested against five alternative models in two large multisource samples (N1 leaders = 537, N1 observers = 7,337; N2 leaders = 1,255, N2 observers = 15,777), each using different leadership instruments. In both samples, the LARI bifactor model outperformed the alternative models. In line with the discrepancy hypothesis, a subsequent variance decomposition showed that each rater source indeed provides unique information about the target’s behavior, although in varying degree. Notably, superiors consistently provided the largest share of unique information among the three observer groups, across all leadership dimensions.