Modeling Psychological Profiles in Volleyball via Mixed-Type Bayesian Networks
Maria Iannario, Dae-Jin Lee, Manuele Leonelli · 2025
Read Modeling Psychological Profiles in Volleyball via Mixed-Type Bayesian Networks by Maria Iannario, Dae-Jin Lee, Manuele Leonelli (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-22111.txt.
Introduction
Maria Iannario, Dae-Jin Lee, and Manuele Leonelli map how mental traits hang together in volleyball players. Their 2025 paper analyzes a new dataset of one hundred sixty-four female players from Italy's C and D leagues, combining standard profiling with background information. The variables are of mixed types: ordered questionnaire scores, categorical demographics, and continuous measures.
To learn directed links among such mixed variables, the authors introduce a hybrid learning method built on a latent Gaussian copula, with a constraint-based skeleton and a score-based refinement. In simulations it recovered network structure more accurately than recent rivals. Applied to the players, the network organized mental skills around goal setting and self-confidence. Emotional arousal linked motivation and anxiety, and personality traits sat upstream of the skill clusters. Scenario analyses show how improving one skill would ripple through preparation, confidence, and self-esteem.
Catalog record
| CREATOR | Maria Iannario, Dae-Jin Lee, Manuele Leonelli |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-22111
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Modeling Psychological Profiles in Volleyball via Mixed-Type Bayesian Networks Maria Iannario, Dae-Jin Lee, Manuele Leonelli Abstract Psychological attributes rarely operate in isolation: coaches reason about networks of related traits. We analyze a new dataset of 164 female volleyball players from Italy’s C and D leagues that combines standardized psychological profiling with background information. To learn directed relationships among mixed-type variables (ordinal questionnaire scores, categorical demographics, continuous indicators), we introduce latent MMHC, a hybrid structure learner that couples a latent Gaussian copula and a constraint-based skeleton with a constrained score-based refinement to return a single DAG. We also study a bootstrap-aggregated variant for stability. In simulations spanning sample size, sparsity, and dimension, latent Max-Min Hill-Climbing (MMHC) attains lower structural Hamming distance and higher edge recall than recent copula-based learners while maintaining high specificity. Applied to volleyball, the learned network organizes mental skills around goal setting and self-confidence, with emotional arousal linking motivation and anxiety, and locates Big-Five traits (notably neuroticism and extraversion) upstream of skill clusters. Scenario analyses quantify how improvements in specific skills propagate through the network to shift preparation, confidence, and self-esteem. The approach provides an interpretable, data-driven framework for profiling psychological traits in sport and for decision support in athlete development. keywords Bayesian networks, Sports analytics, Sports psychology, Structural learning 1 Introduction Elite and amateur sport increasingly recognize that psychological factors are central to performance. Meta-analytic reviews show that key psychological constructs (such as motivation, self-efficacy, affective states, and trained mental skills) are systematically related to competitive outcomes ( Ayranci and Aydin 2025 ; Brown and Fletcher 2017 ; Lochbaum et al. 2021 ) . Personality traits and self-esteem further shape how athletes respond to pressure and to coaching interventions ( Allen et al. 2013 ; Allen and Laborde 2014 ; Laborde et al. 2020 ) . In parallel, the routine collection of standardized psychological questionnaires alongside demographic, training, and performance records has expanded, creating opportunities for data-driven modeling of mental skills in sport ( Albert et al. 2017 ; McGarry 2009 ) . Beyond single-variable analyses, coaches and practitioners typically reason about networks of interrelated traits and contexts ( Laborde et al. 2020 ) . Recent work across both team and individual sports adopts an explicitly multivariate perspective that links personality, coping and mental skills, and performance, and shows that effects frequently operate indirectly through intermediate psychological processes ( Fabbricatore et al. 2023 ; Kalinowski et al. 2020 ; Iannario et al. 2023 ) . Understanding how psychological attributes influence one another, and how they collectively relate to performance, requires models that represent multivariate dependence and conditional structure. Bayesian networks (BNs) provide such a framework: they encode conditional dependencies in a directed graph and support probabilistic reasoning under uncertainty ( Koller and Friedman 2009 ; Lauritzen 1996 ) . In psychology and psychopathology, BNs are now widely used to assess interrelations among traits and symptoms and to probe admissible causal structures from observational data ( Briganti et al. 2023 ; Briganti et al. 2024 ) . In sport, however, particularly for modeling athletes’ psychological profiles, BN applications remain rather limited ( Fuster-Parra et al. 2014 ; Fuster-Parra et al. 2015 ; Ponseti et al. 2019 ) . More broadly across sports analytics, BN methods are still underrepresented relative to regression, SEM, and black-box machine-learning approaches, despite successful demonstrations in outcome prediction, performance assessment, injury prognosis, and tactical decision modeling ( Constantinou 2019 ; Constantinou et al. 2012 ; D’Urso et al. 2023 ; Yung et al. 2025 ; Zhou and Liu 2024 ) . Sport psychology datasets typically combine Likert-scale questionnaire scores (ordinal), demographic or role information (categorical), and training or performance indicators (often continuous) ( Fabbricatore et al. 2023 ; Iannario et al. 2023 ) . In much applied BN work, this heterogeneity is handled by discretizing continuous variables or collapsing categories before analysis. However, ad hoc binning can distort dependence structure, reduce statistical power, and yield graphs that change with the chosen cut points ( Beuzen et al. 2018 ; Nojavan et al. 2017 ) . These limitations have been noted in recent performance-analytics applications as well ( D’Urso et al. 2023 ) . Accordingly, approaches that avoid ad hoc discretization are desirable. There is growing interest in BNs for mixed‐type variables, a setting historically less developed than purely discrete or Gaussian formulations. Such data are precisely the focus of the present work. Classical conditional linear Gaussian models allow discrete and continuous variables but impose parent–child restrictions that limit expressiveness ( Lauritzen and Wermuth 1989 ) . More recent work leverages latent Gaussian copulas with rank-based estimation to model mixed marginals while retaining tractable structure learning ( Hoff 2007 ; Cui et al. 2016 ; Cai et al. 2022 ; Castelletti 2024 ) . Most proposals to date are constraint-based , recovering the graph through sequences of conditional-independence tests that add, remove, and orient edges. Building on these developments, we adopt the “hybrid” strategy familiar from discrete and Gaussian BNs, where a constraint-based phase is followed by a score-based refinement whose search space is restricted by the first phase (e.g., Max–Min Hill-Climbing; Tsamardinos et al. 2006 ). To our knowledge, we are the first to bring this approach to mixed data under a latent Gaussian copula: we extend latent–copula learners ( Cui et al. 2016 ; Cai et al. 2022 ) with a constrained score-based refinement. We term the resulting algorithm latent MMHC . Unlike purely constraint-based copula learners, which may leave some edge directions unresolved, latent MMHC returns a single oriented Directed Acyclic Graph (DAG). We also study a bootstrap-aggregated “stable” variant that repeats the procedure on resamples and aggregates edge-selection frequencies to produce a consensus DAG ( Scutari and Nagarajan 2013 ; Caravagna and Ramazzotti 2021 ) . In simulations spanning variations in sample size, sparsity, and dimensionality, latent MMHC demonstrates improved structural recovery and edge recall compared to competing methods, while maintaining high specificity. We also discuss the associated computational trade-offs and provide open-source R code to facilitate reproducibility. Motivating our methodological development is a dataset on female volleyball players from Italy’s C and D leagues in the Campania region, which combines psychological questionnaires with demographic and experience information. We apply our latent MMHC algorithm and, for robustness, a bootstrap-aggregated stable version to learn an interpretable directed network from these mixed-type variables. The resulting model offers a transparent view of how psychological attributes are conditionally related and provides a basis for broader analyses: identifying traits that may influence others, highlighting potential connectors, and running scenario exercises to explore how improvements in one skill can reverberate through the system. In this way, routine profiling data are transformed into insights that can inform training and athlete development. The paper is structured as follows. Section 2 presents the mixed-type BN framework and introduces the latent MMHC algorithm and its stable variant. Section 3 benchmarks the proposed methods across sample sizes, sparsity levels, and dimensions. Section 4 applies the approach to the volleyball dataset to derive interpretable, coaching-relevant insights. Section 5 summarizes the findings and outlines avenues for future work. 2 Methodology 2.1 Graphical Models and Directed Acyclic Graphs Let \mathbf{X}=(X_{1},\dots,X_{p}) be a vector of random variables. A directed acyclic graph (DAG) \mathcal{G}=(V,E) consists of a