Conference Agenda

A central goal of differential psychology is the development of structural models that organize interindividual construct domains (e.g., personality, abilities, vocational interests, interpersonal behavior. In addition to the traditional zero- and higher-order factor models, there is another class of structural models that has been somewhat overshadowed: circumplex models. Circumplex models are circular structural models that assume a systematic similarity structure among the subdomains that make up a broad construct domain (Guttman, 1954). As such, they have important implications for 1) the interrelationships among subdomains belonging to the same construct domain, 2) multivariate intraindividual profiles within the construct domain, and 3) relationships between such profiles and external variables (i.e., variables outside of the construct domain).

The previous year marked the 70th anniversary of Louis Guttman’s seminal chapter, which laid the foundations for the analysis of circular structures and, thus, for many important theories in differential psychology (e.g., Holland’s RIASEC model of vocational interests or Leary’s interpersonal circumplex), which still play an important role in their respective construct domains. Despite their long tradition, differential psychologists still struggle to use circumplex models to their fullest potential. Moreover, there is a strong trend towards more – not less – complex, specialized, and fine-grained models and analyses that directly contradict the parsimony principle implied by structural models. The aim of this talk is to 1) argue for the continuing relevance of circumplex models, 2) present modern analysis methods and applications, and 3) identify open research questions to be addressed in future studies on this topic.

The Strengths and Difficulties Questionnaire (SDQ) is widely used as a screening tool to assess child and adolescent mental health. When assessing mental health in children and adolescents, measurement invariance (MI) needs to be established to allow for meaningful comparisons across developmental periods. Research on MI of the SDQ across age remains limited, particularly with robust methodological approaches. We used data from the representative German Motorik-Modul Study, testing MI of the SDQ across age groups (3–5, 6–10, 11–13, and 14–17years). We applied multi-group confirmatory factor analysis for ordinal data with weighted least squares estimation, evaluating configural, metric, scalar, and residual invariance. To mitigate overfitting, we employed k-fold cross-validation and validated our findings in a holdout sample. We identified a five-factor model with two residual within factor correlations as the best-fitting structure. Full metric and residual invariance were supported, while partial scalar invariance was established by freeing item intercepts for “worries,” “steals,” “restless,” and “distracted” in specific age groups. K-fold cross-validation confirmed the robustness of these findings. Our results support the use of the SDQ for age-group comparisons while highlighting minor differential item functioning in adolescence. We will extend our findings to compare cross-sectional MI across age groups with longitudinal MI over time. We will also explore machine learning as an alternative to factor analytic approaches to assess MI.

Measurement invariance (MI) analysis on personality questionnaires inevitably runs into several methodological challenges. Using data on the Big Five Inventory 2 (BFI-2, English original and German translation), three of those challenges will be discussed. First, standard confirmatory factor analysis (CFA) models generally do not fit Big Five data well by common standards. This leads to several alternative approaches, each with its own (dis-)advantages. Second, finding a useful criterion for the judgment of MI (and partial MI) in nested CFA models is an ongoing task. It remains difficult to meaningfully define a cut-off between MI and its violation. Third, there is very little research on MI violation indices (i.e., the MI effect size). Nevertheless, judging the degree of MI is crucial to safeguard against a given level of bias in the measure’s application. The analysis of the BFI-2 data will be presented. Methodological decisions on the three outlined challenges will be highlighted, and alternative options will be discussed.

Dissociation is a widespread phenomenon with profound implications for mental health. To date, researchers have relied upon broad conceptualizations and retrospective measures, making it difficult to identify a consistent subset of experiences, explore their links to related phenomena, and differentiate habitual from momentary dissociation. To address these issues, we introduce the Stanford Dissociation Questionnaire (SDQ), a 9-item measure that assesses three facets of dissociation and can be employed to assess either trait or state dissociation. Using the SDQ, we seek to clarify the structure and correlates of dissociation. Across three studies (total N = 976), we demonstrate that trait and state versions of the SDQ exhibit strong content validity and a theoretically congruent factor structure, comprised of alteration in the perception of one’s mind, body, and world (suggesting a tripartite model of dissociation). This structure was largely invariant across demographic groups and time. We also demonstrate good convergent and discriminant validity, internal consistency, and temporal stability over one week. Overall, we conclude that the trait and state versions of the SDQ are reliable and valid tools that can enhance our understanding of dissociation's structure, antecedents, and consequences.