New Developments in Latent Structure Analysis Applied to Social Attitudes

SUMMARY Most analyses based on surveys of social attitudes are limited to the tabulation of response percentages for individual questions and cross-tabulations. In this paper we use methods which seek to extract more information on the pattern of relationships between variables. These are based on the supposition that expressions of attitudes may be explained by assuming that they depend on a few basic dimensions or categories. It is common for response patterns to be incomplete because of refusals or partial responses. This paper describes and illustrates the latent class and latent trait models used in the analysis and gives the appropriate extensions needed to cope with missing values. It also discusses various methods by which the goodness of fit of the models may be judged.