Abstract
This paper introduces a distinct log-linear approach to the analysis of correlates of Guttman-scale positions. I introduce new models, which are referred to as the log-quadratic models. The models are obtained by imposing certain structural constraints on specific quasi-independence models used by Goodman for the Guttman-scale analysis. The derivation of the models owes in part to early studies of Guttman and Suchman on the intensity of attitudes as a structural component of the Guttman scale. The log-quadratic models employ three important structural parameters that characterize (1) the extent of scalability, (2) the mean Guttman-scale position, and (3) the extent of polarization of scale-type response patterns. The third parameter may also be interpreted in certain cases as a measure of attitudinal intensity. These parameters can be used efficiently as dependent variables for a multinomial logit analysis with status covariates. An application examines the effects of region, occupational status, marital status, age cohort and sex on the patterns of responses to a set of questions about mistreatment by tax authorities. A comparison of results with the application of Schwartz's latent-class models to the same data is also made.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call