The proposed project deals with extensions of both theoretical and applied aspects of latent class analysis, a statistical technique for subdividing sets of data on individuals into groups whose properties are relevant to the subject of scientific inquiry. In survey research, for example, it is difficult to say how many U.S. citizens are in favor of, opposed to, or indifferent toward some particular issue under investigation unless we ask different kinds of questions related to that issue and then take account of the many possible ways that people actually respond to those questions. Latent class analysis is one statistical method for getting around this measurement problem. It allows the researcher to classify the population (as estimated by some sample) into groups which summarize all the ways individuals might respond to the various survey items. Typical uses of the method include the measurement of abilities or aptitudes in the field of education research (to assist in the selection of test questions, for instance), and the analysis of respondent data obtained in surveys of attitudes, expections, or preferences in public opinion polling or in the field of market research (to help asses the extent of change over time, for instance). The goal of the present project is to extend the applicability of latent class analysis by building general- purpose procedures that can be employed in areas of the social sciences.
