Ires Ghielen

65 Replication factor analysis in neuropsychiatric PD patients Scoring of motor symptoms, using the UPDRS-III [16], was performed on a four-point Likert scale, ranging from 0 to 3 per item. The UPDRS-III contains 39 items and a higher score represents more severe motor symptoms. Symptoms of anxiety were measured with the BAI. The BAI is a 21-item self-report instrument asking for symptoms of anxiety over the past week [13]. Patients answer on a four-point Likert scale, ranging from 0 (not at all) to 3 (severely) per item. Statistical analyses The analyses were performed in the statistical program R (version 3.6.1, developed by the R Core Team, 2015). Differences in descriptive measures were investigated using independent t-tests or Mann-Whitney U tests, and chi-square tests where appropriate. The package qgraph was used to perform the network analyses and visualizations [14]. Since it concerned ordinal data, a Gaussian Graphical Model with LASSO regularization was applied [15]. The tuning parameter, gamma , was set at 0.5 to obtain a network structure in which few associations are required to parsimoniously explain the covariance among the variables, reducing false positive associations [15]. Associations within a network are represented by color-coded lines between symptoms, which are green for positive associations and red for negative associations. The thicker the lines, the stronger the association between two symptoms. The Network Comparison Test (NCT) was used to investigate differences in associations between motor and anxiety symptoms between the two patient groups [16]. From the NCT, two types of measures were extracted. The first measure concerns global strength, which represents the overall connectedness between all symptoms within the network. A higher global strength results from more and stronger associations between symptoms. Second, all associations between separate symptoms are calculated individually, and can be compared between the two patient group networks. The Holm-Bonferroni method was used to correct for multiple comparisons. To investigate the accuracy and stability of our network estimates, a bootstrap procedure was applied. Using the R package bootnet , a number of 100 bootstrap samples with the nonparametric method was used [15]. The sampling distribution was then visually inspected. 4