Taller 10. Log-lineal models
25 Julio 2016 10:00 am - 29 Julio 2016 2:00 pm
El Colegio de México, Ciudad de México
Log-linear models are for analyzing relationships among categorical variables; they are widely used in studies of assortative mating, inter-generational class mobility, and educational attainment. They can be useful when traditional regression-based approaches are problematic. The workshop introduces the basic principles of research design using the log-linear approach. Students will learn how to conceptualize, estimate, interpret, and present the different types of models, including topological, linear-by-linear, hybrid, log-multiplicative, and other extensions.
Resumen: Log-linear models are for analyzing systems of relationships among categorical (discrete) variables, both ordered and unordered. They are widely used in studies of assortative mating, inter-generational class mobility, and educational attainment, and they have potentially broad applications. They are especially useful for the analysis of variables with numerous response categories. The course begins by presenting the basic logic of log-linear analysis, including statistical assumptions, alternative parameterizations, principles for making inferences and assessing model fit, and parallels with more familiar regression analysis. Then we will examine different types of log-linear models, including topological models, linear-by-linear models, association models, hybrid models, log-multiplicative models, and various extensions of these. Students will learn the basic principles of research design using the log-linear approach, the tools for estimating the different types of models, and principles for assessing different models, interpreting results, and presenting findings. Readings will include both methodological texts and substantive articles drawn from family demography and stratification studies. Sessions will consist of lectures, lab time for software instruction (including both Stata and the special package, LEM, which is set up to easily estimate a wide range of log-linear models), short hands-on software assignments, and discussions. The discussions will focus on the methodological rationale underlying the substantive claims made in articles on assortative mating and inter-generational class mobility. Students who complete the course will have a basic familiarity with a wide range of log-linear models and the ability to design a research project using a log-linear approach, estimate the models and interpret results, and critically assess studies that use these techniques.