Abstract
A structural equation model for panel designs. In the fields of social psychology and sociology of attitudes, the data obtained from longitudinal panel designs have been analyzed mainly by structural equation systems of each model: a) using first level latent variable models, in which several latent variables represent the content of each variable observed at different times, while other latent variables represent each wave of measurement (Jagodzinsky, Kühnel and Schmidt, 1987, 1988, 1990); b) other models analyze the same designs by means of 'true scores' of observed variables (Saris and Van den Putte, 1988). In this paper a secondary analysis is made of a 4V3W design, concerning attitudes towards guest workers using the two previously cited models as well as a second level latent variable model which represents the general attitude towards guest workers. This second level latent variable influences each measurement wave and each first level latent variable, representing each observed variable. Results indicate that this latter model fits the data very well (in spite of being theoretically and empirically less parsimonious than the former models). The advantages of a second level latent variable model are that it has greater scope for extension and, from a cognitive perspective, it is more suited to the behavior of the subject.
