Eilert Sundts hus
4th floor (map)
Moltke Moesvei 31
Lecturer: Professor Melinda Mills,
Department of Sociology,
University of Groningen, Groningen,
Main disciplines: Sociology, Demography, Statistics
Dates: 2 - 6 August 2010
Course Credits: 10 pts (ECTS)
Limitation: 30 participants
Content and objectives of course
The life course approach is a means to study patterns and trajectories of individuals. The aim of this course is to provide an overview of theoretical and methodological approaches to cross-national life course research, with a focus on applied event history analysis.
By the end of this course, students will be able to:
The first days of the course will introduce students to the basic concepts and central topics in life course research. We will then move from theory to empirical applications, focussing on cross-nationally comparative life course research and related studies of inequality. The remainder of the course will introduce students to the central methodological approach in life course research, survival and event history analysis. Here the focus will be on practical computer-based instructions of survival and event history analysis using the computer programme R.
For the first day of the course, students should prepare a brief introduction of their own (planned or actual) individual research project. The introduction should be very brief and no longer than 3 minutes (i.e., it is a verbal introduction, no power point presentations, etc.).
Students also have the option or writing a 6,000 to 10,000 word research paper within eight weeks after the course to receive a course certificate and earn credits. The topic of the paper should be decided in consultation with the instructor. Students may focus on a content-related essay or undertake a related detailed analysis using their own data. A more detailed specification of this essay and the criteria for grading will be provided at the beginning of the course. Students who fulfil this requirement with a passing grade will receive 10 ECTS credits.
Essential readings for preparation of the course
Students should obtain and read these books/articles in advance of the course.
LECTURE OUTLINE (and required readings)
This course consists of approximately 4 lecture-hours each day, divided into a morning and afternoon session, each consisting of two 45 minute lectures.
Monday 2 August: Introduction, life course theory and empirical applications
Introduction: Brief introduction and discussion of (planned or actual) individual research projects and designs (each student should prepare a short 3 minute verbal introduction of their research topic/interests)
Lecture 1: Life course research
The goal of this lecture is to introduce students to the main theories, concepts and challenges in contemporary life course research.
Lecture 2: Cross-national applications and inequalities in the life course
During this lecture, students will learn about cross-national comparative life course and inequality research and actively discuss applications and challenges.
Tuesday 3 August: Introduction to Event History Modelling of the Life Course and R
Lecture 3: Introduction to event history modelling of the life course and data
The goal of this lecture is to introduce students to the fundamental concepts and terminology in event history analysis, understand why these methods are useful and which problems they can solve, understand censored and truncated data, survivor and hazard function and their relationship and have a general overview of the different types of models and classes. In the second part of this lecture, students will gain insight into different types of data structures including the single and multi-episode files, person-period files with and without lagged variables and episode-splitting.
Lecture 4: Introduction to R and your first session
The goal of this lecture is introduce students to different computer programmes for modelling the life course and provide a basic introduction to R. Students will then learn how to load the related packages in R for these types of analyses, open and examine data, run basic descriptives and summary statistics and graphical explorations of the data.
Wednesday 4 August: Nonparametric Methods and the Cox Model
Lecture 5: Nonparametric methods
The goal of this lecture is to allow students to: understand the basic tenants and calculations of the Kaplan-Meier (KM) estimator, conduct and interpret KM analyses in R, produce a univariate KM plot, plot two KM curves to compare survival between groups, determine whether differences are statistically significant between groups and understand why it may be useful to stratify the analysis.
Lecture 6: The Cox-proportional hazards regression model
The goal of this lecture is to allow students to be able to: recognize the general and time-varying form of the Cox proportional hazards model and understand why it is useful, and understand the meaning of the proportional hazards assumption. Students should also be able to estimate and interpret a Cox regression model with fixed and time-varying covariates, interpret the hazard ratio, test the significance of the model, produce and interpret survival curves, integrate time-varying covariates by producing a person-period file, reduce the problem of causal ordering by introducing lagged time-varying covariates and examine time-dependence.
Blossfeld, H.-P. and M. Mills (2001) A Causal Approach to Interrelated Family Events: A Cross-National Comparison of Cohabitation, Nonmarital Conception and Marriage,’Canadian Studies in Population, 28(2): 409-437. (28 pages)
Mills, M. (2010) The Cox Proportional Hazards Regression Model, Chapter 6, Introducing Survival and Event History Analysis, London: Sage. (28 pages)
Thursday 5 August: Parametric Models, Models Building and Diagnostics
Lecture 7:Parametric models
The aim of this lecture is to first introduce students to the main characteristics of parametric models and why you should use them and the difference between accelerated failure time (AFT) versus proportional hazards (PH) models. Students will then learn how to estimate and interpret a selection of AFT and PH model specifications (e.g., exponential, Weibull) and then reshape the data to estimate and interpret a piecewise constant exponential model.
Lecture 8: Model building and diagnostics
The aim of this lecture is allow students to understand the cumulative research process of building an appropriate model, understand the importance of a purposeful selection of covariates and assess the overall goodness of fit of models in order to choose an appropriate model. Students will then learn how residuals can be used to evaluate a model to: test overall model adequacy, check for a violation of the proportional hazards assumption, deal with non-proportional hazards, check for influential observations and detect non-linearity.
Friday 6 August: Multilevel Models, Multiple Events and Entire Histories
Lecture 9: Multilevel and frailty event history models for cross-national comparative research
The aim of this lecture is to introduce students to the problem of correlated data, understand the analysis of recurrent events and frailty and recognize different forms of frailty. Students should also have a basic ability to estimate and interpret Cox and parametric frailty (multilevel) models.
Lecture 10: Modelling multiple events and entire histories: competing risk, multistate models and sequence analysis
The goal in this lecture is to first introduce students to the analysis of competing risks and the central techniques used to model them (latent, cumulative incidence curve (CIC)). Students will then briefly learn about how to prepare data, estimate, compare and interpret competing risks models. Another goal is that students have a basic understanding of multistate models and their applications and could prepare data, estimate and interpret these models. Students will then be given an introduction to the basics of sequence analysis of entire trajectories and shown how to prepare, describe and visualize sequence data, estimate and interpret the similarities and distances between sequences using the optimal matching approach, engage in a cluster analysis to produce prominent typologies of sequence trajectories, and estimate and interpret basic event sequence analysis results.
Question and answer session: Discussion of individual research projects and designs.
Melinda Mills (PhD Demography 2000) is Professor of the Sociology of the Life Course at the Department of Sociology, University of Groningen, Netherlands. Since 2003, she has been the Editor of International Sociology (official journal of the International Sociological Association). She has written numerous articles and several books including: Introducing Survival and Event History Analysis (2010, London: Sage), the co-edited books Globalization, Uncertainty and Youth in Society (2005, London: Routledge), and Globalization, Uncertainty and Men’s Careers: An International Comparison (2006, Cheltenham: Edward Elgar). In 2009, she was the guest Editor of a special issue for the European Sociological Review, on Globalization and Inequality.