ABSTRACT
The purpose of this chapter is to demonstrate data analytic techniques designed to account for the measurement, distributional, and hierarchically structured properties of data obtained in organizational psychological research. Whereas there is an abundance of literature advocating what should or should not be done in analyzing psychosocial data to address specific research questions, there is all too little that provides practical, illustrative examples. Notable exceptions for applications in structural equation modeling include the texts by Cuttance and Ecob (1987), Marcoulides and Schumacker (1996), and by Schumacker and Lomax (1996). For applications in multilevel analysis, the texts by Bryk and Raudenbush (1992); Goldstein (1987, 1995); Hox (1994); Kreft and de Leeuw (1998); and the users’ guides by du Toit and du Toit (1999), Goldstein et al. (1998) and Rasbash et al. (2000a) are useful, as are the workshop manuals by du Toit, du Toit, and Cudeck (1999), and by Rowe (1999a, 1999b, 2001). Thus, using a data set designed to explain variation in teachers’ cognitive/affective constructions of their roles and perceptions of their work environments, this chapter illustrates the use of one-factor, congeneric measurement models to obtain maximally reliable composite variables, and the utility of multilevel analytic techniques in fitting regression and structural equation models to estimate the magnitude of the interdependent effects among those variables.
