Comparing factor analysis and the Rasch model for ordered response categories: an investigation of the scale of gambling choices

Abstract

Using both factor analysis (Spearman, 1904) and the Rasch model for ordered response categories (Andrich, 1978), the present study investigated the structure of the Scale of Gambling Choices (SGC, Baron, Dickerson and Blaszczynski, 1995). The scale was administered to a participant sample (n = 210) consisting of 57 first year psychology students, 104 in situ club Electronic Gaming Machine (EGM) players and 49 self-referred problem gamblers. It was hypothesised that the results yielded by factor analysis and Andrich's model would not agree with respect to the behaviour of individual items. This hypothesis was supported; supporting the results of previous research (Johnson, et al., 1995; Raju, et al., 2002; Reise, et al., 1993). It was also hypothesised that a relationship would exist between item factor loadings and item expected value curve slope coefficients. This hypothesis was not supported and so hence did not support the findings of Parsons and Hulin (1982) and Roskam (1985). It was concluded that this was perhaps due to the different latent variable conceptions which exist between the Rasch models and factor analysis (Bollen, 2002). The limitations of the research were outlined and suggestions for future research were made.