This course is an introduction to theory and applications of event-history analysis, plus some elements of panel data analysis. Longitudinal data are commonly used to address many research questions in demography, social sciences, and epidemiology.
- Blossfeld, H., K. Golsch, and G. Rohwer. 2007. Event History Analysis with Stata. Mahwah, NJ: Lawrence Erlbaum.
- Scherer, S. 2013. Analisi dei dati longitudinali. Un'introduzione pratica. Bologna: Il Mulino.
- Slides and additional materials will be provided during the course.
Learning Objectives
This course covers univariate and multivariate (regression) methods for analysis of duration (event-history) data, including their recent developments. Students also learn data management skills that are specific to conducting event-history analysis in Stata.
Finally, students will be able to discuss applications in the domain of social and demographic research.
Prerequisites
Statistical inference.
Teaching Methods
Face-to-face lessons and lab sessions.
Type of Assessment
Written exam that icludes both exercises in Stata and questions on theory. The final evaluation will take into account student's discussion of a research article applying longitudinal data analysis (up to 4 points on the final mark).
Course program
Introduction (Basic concepts and definitions, Event history data, censoring and truncation, discrete vs. continuous time); Event history data (Coding and data preparation, Life tables, Kaplan-Meier, related estimators, Stata applications, time-constant and time-varying variables); Non-parametric models (Exponential and piece-wise constant models); Modelling-related issues (Interactions and combinations of variables; model choice and goodness of fit); Parametric models (Weibull, Gompertz, Log-Logistic, Log-Normal); Cox model (Estimation, interpretation of parameters and model diagnostics, PH assumption); Competing risk models (Data preparation, estimation and interpretation); Advanced topics (Discrete time models, frailty models – unobserved heterogeneity); Introduction to panel data and related regression models (random and fixed effects).