Elite Athletes Data Analysis

<p>This dataset contains the R script that was used for the network analysis of sociodemographic, sport-related, medical and psychometric data of elite athletes data. The aim of the study was to assess data on mental health of elite athletes and investigate associations and interconnections among different variables using network analysis. Data was collected through a digital cross-sectional study. The survey collected sociodemographic data, including financial situation. Medical data covered body height, body mass, medications, and injuries within the last 12 months (regardless of type, location, and whether or not it was a contact injury). Since the survey addressed elite athletes, it also covered different sport-related data such as type of sports, years in elite sports, number of training units per week, duration of training units, etc. Moreover, five validated measures were used in the survey to assess aspects of mental health symptoms, namely generalized anxiety symptoms, depressive symptoms, somatic symptom disorder symptoms and psychological distress.</p>
<b>Methods</b>
<p>Network Analysis was performed using the packages qgraph, igraph, bootnet, and EGAnet (Csardi & Nepusz, 2006; Epskamp et al., 2012; Golino & Epskamp, 2017). Centrality indices were computed and assessment of the network's stability and accuracy was conducted via bootnet. Missing data was addressed using listwise deletion, with the minimum sample size set to 250-350 participants to ensure sufficient power for the analysis of networks with 20 nodes or fewer (Constantin et al., 2021). The study estimated and visualized the network using a gaussian graphical model (Epskamp & Fried, 2018). Depressive symptoms, somatic symptom disorder, generalized anxiety, distress, mild to moderate injuries, severe injuries, years in elite sports, substance use, financial situation and training units per week were selected as nodes, resulting in a total of 11 nodes in the network. The dependencies among the variables were represented as edges in the network based on partial correlations (Epskamp & Fried, 2018). According to Epskamp and Fried (2018), gLASSO and EBIC (Chen & Chen, 2008; Friedman et al., 2008) methods were applied, with a tuning parameter of 0.5. The tuning parameter of 0.5 was chosen to create a parsimonious network with a higher specificity, as suggested by Epskamp and Fried (Epskamp and Fried, 2018). The centrality indices were then calculated to determine the importance of each node in the network. These indices included degree centrality, strength, closeness, and betweenness (Hevey, 2018). Degree centrality is the sum of all edges of a node, strength is the sum of the edge weights of all edges of a node, closeness measures the average distance of a node to other nodes, and betweenness identifies the role of a node in connecting other nodes (Hevey, 2018). The centrality indices are intended to provide clues as to which constructs are particularly relevant in the context of various mental health and sport variables (Epskamp and Fried, 2018). The stability and accuracy of the network were evaluated through different bootstrap procedures, including an edge weight variation analysis (Isvoranu et al., 2021) and a correlation stability analysis. It is recommended that, in order to interpret centrality with confidence, stability coefficients should exceed at least .25 and ideally surpass .50 (Epskamp et al., 2018a) . The interpretability of the edge weight, node strength, and centrality indices was also assessed.</p>