There is (still) a global factor that underlies the PGSI: A reanalysis of Tseng, Flack, Caudwell, and Stevens (2023)

• Published in: Addictive behaviors Vol. 140; p. 107623
• Main Authors: Tabri, Nassim, J. A. Wohl, Michael
• Format: Journal Article
• Language: English
• Published: England Elsevier Ltd 01-05-2023
• Subjects: Australia - epidemiology; Behavior, Addictive; Disordered Gambling; Factor Analysis, Statistical; Gambling; Humans; Measurement; Problem Behavior; Problem Gambling Severity Index; Psychometrics; Severity of Illness Index; Structural Equation Modelling; Australia; Measurement; Psychometrics; Problem Gambling Severity Index; Disordered Gambling; Structural Equation Modelling
• ISSN: 0306-4603; 1873-6327
• DOI: 10.1016/j.addbeh.2023.107623

•Tseng et al. (2023) argued that 2 factors underlie scores on the PGSI items.•In our reanalyses a hierarchical model—1 global & 2 sub factors—best fit their data.•The global factor was empirically well-defined (all items had strong loadings)•The two sub factors were not empirically well-defined (weak & medium loadings)•Continued use of the PGSI total score in research and assessment is recommended.
Tseng, Flack, Caudwell, and Stevens (2023) conducted confirmatory factor analyses (CFA) of the Problem Gambling Severity Index (PGSI)—a gold-standard measure of disordered gambling symptomatology that has traditionally been indexed using a total score—with data from a large representative sample of Australians residing in Northern Territory who gamble (N = 3,740). Based on their results, Tseng et al. argued that a two-factor model best fit the data and so the PGSI items could be separated into two subscales: Problem Behaviours and Consequences of Problem Behaviours. We reanalyzed their data using CFA and found that a hierarchical model provided the best fit to the data. The hierarchical model includes a global factor underlying all PGSI items and two sub factors that correspond to the PGSI items assessing behaviours and consequences. The global factor was empirically well-defined, but the behaviours and consequences sub factors were not. Also, the two sub factors did not reliably measure the more narrowly defined behaviours and consequences constructs independent of the global factor. Based on our reanalyses of Tseng et al.’s (2023) data, we encourage researchers in the field of gambling studies to continue using the PGSI total score as an index of disordered gambling symptomatology.