Search results

Identifying the dimensionality of a construct and selecting appropriate items for measuring the dimensions are important elements of marketing scale development. Scales for measuring marketing constructs such as service quality, brand equity, and marketing orientation have typically been developed using the influential classical test theory paradigm (Churchill, 1979), or some variant thereof. Users of the paradigm typically assume, albeit implicitly, that items and respondents are the only sources of variance and respondents are the objects of measurement. Yet, marketers need scales for other important managerial purposes, such as benchmarking, tracking, and perceptual mapping, each of which requires a scaling of objects other than respondents such as products, brands, retail stores, websites, firms, advertisements, or social media content. Scales that are developed without such objects in mind might not perform as expected. Finn and Kayande (2005) proposed a multivariate multiple objective random effects methodology (referred to here as M-MORE) could be used to identify construct dimensionality and select appropriate items for multiple objects of measurement. This chapter applies M-MORE to multivariate generalizability theory data collected to assess online retailer websites in the early 2000s to identify the dimensionality of and to select appropriate items for scaling website quality. The results are compared with those produced by traditional methods.