On successful completion of this course, students will be able to...
- describe and apply the main tools for likelihood-based inference with categorical variables, such as maximum likelihood (ML) estimation, computation of SEs and confidence intervals, Wald, Likelihood-ratio, and Score tests, and the delta method
- apply the generalized linear modeling framework, including its inferential and computational tools, and explain the underlying statistical theory
- explain and apply the different types of models for nominal and ordinal response variables
- describe, apply and evaluate categorical data techniques for dependent observations, such a latent class analysis and generalized linear mixed models
- understand, evaluate, and synthesize scientific articles on statistical methods for the analysis of categorical data
The most important theme in this course is regression analysis with non-continuous dependent variables using binary, multinomial and ordinal logistic, probit, and Poisson regression models; that is, regression models belonging to the generalized linear modelling family. In addition, we go into the details of technical aspects such as the principle of and algorithms for maximum likelihood estimation and likelihood-based statistical inference (SE and CI computation, Wald, score, and likelihood-ratio testing; and delta method). The third topic is latent class analysis, which is a method for cluster analysis, scaling, and the random effects modeling. The relevant software packages in this course are SPSS, R, Latent GOLD, and Excel.
- Agresti, A., Categorical data analysis, Hoboken, NJ: Wiley, 2002 (second edition), ISBN 0-4713-6093-1.
- Magidson, J., and Vermunt, J.K., ( 2004) Latent class models. D. Kaplan (ed.), The Sage Handbook of Quantitative Methodology for the Social Sciences, Chapter 10, 175-198. Thousand Oaks: Sage Publications. (update version)
- Research papers to be handed out during the course.