Quantitative Methods in the
Social Sciences 2

Summer School on Analysing the life course

Advanced methods for the analysis of complex event history data

Summary:

Following on from the Paris summer school in 2009 which provided an introduction to event history analysis, the Bristol summer school will cover more advanced topics with a focus on methods for analysing interdependent events (e.g. women’s employment and childbearing). The summer school will consider descriptive techniques for exploring complex histories as well as statistical modelling approaches.

Sequence analysis (2 days). Dr Alexis Gabadinho and Dr Matthias Studer (Laboratoire de démographie et d'études familiales, University of Geneva) are two of the developers of the TraMineR software. The course is devoted to the analysis of state or event sequences describing life trajectories such as family life courses or employment histories. The course will be practically oriented, including an introduction to the R statistical environment and training in the TraMineR library for mining and visualizing sequences. The following topics will be covered:

• Longitudinal data formats and converting between various sequence representations
• Describing and visualizing sequences data sets
• Measuring sequence complexity
• Building sequence typologies using clustering
• Heterogeneity measures and ANOVA-like analysis of sets of sequences
• Mining event sequences

Modelling recurrent events and transitions between multiple states (1 day). Most lifecourse events can be experienced more than once by an individual, and recurrent events can be viewed as a having a hierarchical structure with multiple ‘at risk’ episodes nested within individuals. Further complexity arises when the occurrence of an event may mark the transition to a different state (e.g. formation of the first union marks a transition from the ‘single’ to the ‘marriage’ or ‘cohabitation’ state). Multilevel (random effects) models for recurrent events and multiple states will be considered.

Event history analysis with endogenous predictors (2.5 days). Many covariates we consider in event history analysis are potentially endogenous, that is jointly determined with the timing of the event of interest. For example, in an analysis of the impact of childbearing on women’s employment, time-varying indicators of the presence and number of children are outcomes of a process that may have similar unmeasured determinants to employment transitions. Another example where endogeneneity is a particular concern is in evaluations of selective interventions. The course will introduce methods that allow for such dependencies, for example simultaneous equation (multiprocess) models.

Prerequisites:

As this is an advanced course, knowledge of standard event history methods for analysing the time to a single event will be assumed.
Before attending the summer school, participants should ensure that they are familiar with the methods described in:

Blossfeld, Hans-Peter, Golsch, Katrin and Rohwer, Götz (2007). Event History Analysis with Stata. Lawrence Erlbaum Associates. (Chapters 1-7 and 9)

No prior experience with any particular statistics package will be assumed.