UV9257U – Multilevel and Longitudinal Modeling
Course content
This course has been replaced with UV9257 Multilevel and Longitudinal Modeling.
This course will be offered for the last time in the spring of 2022. The course will be held on 30th May-2nd June 2022.
Welcome to this 4 day PhD-course lead by Professor Sophia Rabe-Hesketh, University of California, Berkeley, and Professor Anders Skrondal, CEMO & Norwegian Institute of Public Health & University of California, Berkeley.
The short course introduces models for multilevel or clustered data, such as cross-sectional data with students nested in schools, or longitudinal data with repeated measures/panel waves nested in subjects. Models and concepts are introduced via examples from a variety of disciplines, equations and illustrative graphs, keeping the mathematics as simple as possible (avoiding matrix algebra and calculus). However, the course covers many topics and may be conceptually demanding. Software is not discussed, but the handouts include Stata commands for all the results that are presented and the data are available online. The short course is based on successful semester-long graduate-level courses by the presenters at Berkeley and London School of Economics.
Literature:
- Rabe-Hesketh, S. and Skrondal, A. (2012). Multilevel and Longitudinal Modeling Using Stata (3rd Edition). College Station, TX: Stata Press →Vol. 1 on “Continuous Responses” is sufficient together with Chapter 10 on “Dichotomous or Binary Responses” from Vol. 2: http://www.stata-press.com/books/mlmus3_ch10.pdf
- Snijders, T.A.B., and Bosker, R.J. (2011). Multilevel Analysis. An Introduction to Basic and Advanced Multilevel Modelling (2nd Edition). London, Sage.?
Learning outcome
Part 1 of the course covers linear multilevel models for continuous responses, including random-intercept, random-coefficient, and three-level models. (Restricted) maximum likelihood estimation of model parameters and empirical Bayes prediction of random effects are introduced at a non-technical level. Part 2 focuses on longitudinal data analysis, starting with application of random-coefficient models for growth, contrasting this approach with marginal modeling and giving a brief overview of methods from panel data econometrics. Part 3 introduces multilevel logistic regression for binary responses.
By the end of the course, they should have an understanding of the model assumptions, be able to choose an appropriate model for a given situation and interpret the parameter estimates.
Admission
Ph.d.-students from the Faculty of Educational Sciences at the University of Oslo apply through Studentweb.
All others can apply through Nettskjema.
Registration deadline: 30 April 2021
Limited number of Places
Prerequisites
Recommended previous knowledge
Prior to the course, participants must be familiar with linear and logistic regression.
Teaching
Outline of Topics
Part 1: Linear mixed models
-
Variance-components models
-
Random-intercept models
-
Predicting random effects
-
Random-coefficient models
-
Three-level models
Part 2: Longitudinal data analysis
-
Longitudinal data
-
Growth curve models
-
Marginal models
-
Panel data econometrics
Part 3: Multilevel models for binary data
-
Review of multiple logistic and probit regression
-
Multilevel logistic models
-
Estimation and prediction
To see the schedule, please see the semester page.