Content

The event-history module starts with one of the earliest methods for the analysis of longitudinal data available: the life table. The life table provides a convenient means to introduce the concepts and terminology used in longitudinal analysis: events, risk sets, state space, duration intervals, person-periods, probabilities and transition rates. The life table subsequently shows data on date-of-event-occurrence or age-at-event-occurrence readily encountered in survey material can be converted to longitudinal duration data on event-occurrence. The measurement of duration data is a key element distinguishing different types of event-history analysis. Both discrete-time methods and semi-parametric or continuous-time methods are developed throughout the module.

The complexity of event-history analysis is gradually increased throughout the module. The module starts out with the simplest case: the analysis of a non-repeatable event of a single kind (e.g. death, first job…). For discrete-time event-history analysis the model of logistic regression is treated first as it offers a straightforward means of analysing transition probabilities as a function of covariates when used in combination with a person-period file. The logit model of the discrete-time hazard is subsequently compared with models using alternative link functions, such as the complementary log-log link function. The discrete-time models are subsequently compared to parametric methods for continuous-time data and the (semi-parametric) proportional likelihood model introduced by Cox. Following the analysis of non-repeatable events of a single kind, more complex transitions are considered. A research problem that is frequently encountered in the analysis of longitudinal data is the occurrence of multiple kinds of events (eg. death from different causes, differentiation of the entry into the labour market by type of job…). Multinomial logit models are introduced to analyze competing risks in discrete-time. The concept of ‘type-specific hazard functions’ is introduced on the side of continuous-time models. Subsequently, the basic types of event-history analysis discussed for the analysis of events of a single kind and multiple kinds of events are applied in more complicated ways to analyse repeated events and changes of states. Finally, the module considers frailty and shared frailty models in both discrete- and continuous-time.

Throughout the module, the different types of models are applied to real-world data on the transition to employment among Turkish and Moroccan immigrants in Belgium, data from the National Databank Mortality on socio-economic mortality differentials as well as data on labour force participation and family formation drawn from retrospective surveys such as the Fertility and Family Surveys (FFS), the European Social Survey (ESS) and the Generations and Gender Survey (GGS).
 

Course material

Suggested readings:

• Allison, P.D. (1982) Discrete-time methods for the analysis of event-histories, In: Leinhardt S. (ed.) Sociological Methodology, San Francisco, Jossey-Bass, P. 61-98.
• Allison, P.D. (1984) Event History Analysis. Regression for Longitudinal Event Data. Sage University Papers on Quantitative Applications in the Social Sciences, 07-001, Beverly Hills and London, Sage Publications.
• Allison P.D. (2004) Event-history Analysis, In: Hardy M., Bryman A. (eds.)(2004) Handbook of data analysis, London, Thousand Oaks, New Delhi, Sage Publications, 369-385.
• Blossfeld, H.-P., K. Golsch and G. Rohwer. (2007) Event History Analysis with Stata. London, Lawrence Erlbaum Associates.
• Cleves, M. A., Gould W. W., Gutierrez. R. G. (2010) An Introduction to Survival Analysis using Stata. Third Edition., College Station: Stata Press.
• Hinde A. (1998) Demographic Methods, London, Arnold.
• Menard, Scott (1995) Applied Logistic Regression Analysis.Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-106. Thousand Oaks, CA: Sage.
• Mills M. (2011) Introducing Survival and Event History Analysis, London, Sage Publications.
• Putter, H., M. Fiocco, and R.B. Geskus, Tutorial in biostatistics: Competing risks and multi-state models. Statistics in Medicine, 2007. 26(11): p. 2389-2430. 
• Rizopoulos D. (2012) Joint Models for Longitudinal and Time-to-event Data: With Applications in R, Chapman & Hall, CRC Biostatistics Series.
• Singer J., Willett J. (2003) Applied Longitudinal Data Analysis. Modeling Change and Event Occurrence, Oxford, Oxford University Press.
• Wienke A. (2010) Frailty Models in Survival Analysis, Chapman & Hall, CRC Biostatistics Series.
• Yamaguchi, K. (1991) Event History Analysis, Sage Applied Social Research Methods Series, volume 28, Thousand Oaks, CA: Sage.

Language of instruction: more information

All theory classes and applied computer lab sessions are taught in English.

Format: more information

Computer session

The module consists of both theory classes and applied computer lab sessions- with hands-on exercises in class (using spreadsheets and statistical software). In the theory classes the different types of models in discrete-time and continuous-time are introduced and subsequently applied to real-world data. In the applied computer lab sessions students are trained to estimate the difference types of models discussed previously in the theory classes with applications in SPSS, Stata and R.  

This course module is taught in block teaching. More specifically, the lectures and computer class sessions will take place in 4 subsequent weeks in the first semester (multiple classes a week). Screen captures (slides + audio) will be made available after each session.