23-24 June, 2010 by Andrew Pickles (University of Manchester)
The course describes how an extensive range of models can be seen as special cases of the generalized linear latent and mixed model and illustrates this with a series of examples using the Stata procedure GLLAMM. These models include confirmatory factor analysis, random effects, structural equation and mixture distribution models for a range and a mixture of response data types. The course will end with a consideration of more novel model types, such as multilevel structural equation and non-ignorable non-response models.
Day 1: GLLAMM framework & longitudinal data:
- GLLAMM framework & random intercept multilevel models
- Fixed effects, random effects, analysis of covariance and the unified longitudinal model
- Random coefficients and growth curve models
Day 2: Measurement, trajectory and causal models:
- Factor and IRT models
- Latent Class & trajectory models
- Causal analysis, instrumental variables and non-response models
On both days there will be combination of lectures and lab sessions
Participants should be familiar with regression, logistic regression, and the construction and use of dummy variables for categorical predictors. Though an introduction to Stata is provided it is extremely brief and prior experience in the use of Stata allows participants to focus more on the modelling issues.
Andrew Pickles is Professor of Epidemiological and Social Statistics at the University of Manchester and is shortly to take up the post of Professor of Biostatistics and Psychological Methods at the Institute of Psychiatry, King’s College. His research interests include structural equations modeling and longitudinal data analysis. The context of much of his work has been longitudinal behavioral and developmental psychopathological studies in which we attempt to understand the interplay between social, psychological and biological processes . Other studies have included voting behaviour and social capital. Together with Sophia Rabe-Hesketh and Anders Skrondal he formulated the generalized linear latent and mixed model (gllamm) implemented in Stata.