|
|
Legend:  = Applied Session,
 = Theme Session,
 = Presenter
FRY = Fairmont Royal York, ICH = InterContinental Hotel, TCC = Metro Toronto Convention Center
|
|
|
CE_02C
|
Sat, 8/7/04, 8:00 AM - 4:00 PM
|
FRY-Manitoba
|
|
Generalized Linear Latent and Mixed Models (1 Day Course) - Continuing Education - Course
|
|
ASA, Biometrics Section
|
|
Instructor(s): Anders Skrondal, Norwegian Institute of Public Health, Sophia Rabe-Hesketh, University of California, Berkeley, Andrew Pickles, University of Manchester
|
|
Generalized linear mixed (or multilevel) models are useful for longitudinal data, cluster-randomized trials, surveys with cluster-sampling, genetic studies, meta-analysis and many other applications. The random coefficients in generalized linear mixed models are latent variables representing between-cluster variability and inducing within-cluster correlations. Latent variables are also often used to represent true values measured with error, a typical example being diet (continuous latent variable) or diagnosis (categorical latent variable). Measurement models relating measured variables to latent variables can be used to investigate the properties of measurement instruments or diagnostic tests. Such measurement models can also form part of structural equation models relating latent variables to other latent or observed variables. An important application is regression with covariate measurement error. Finally, latent variables can be used to model the dependence between different processes, for instance the response of interest in a clinical trial and the (non-ignorable) drop-out process.
All these models have very a similar structure. However, this is not commonly recognized due to disparate terminologies and lack of communication between methodologists in different disciplines, for instance (biostatisticians, econometricians and psychometricians. Taking a unified view is beneficial since developments for one model-type are often applicable to other model-types. Furthermore, the same software can often be used to estimate seemingly different models.
The course will be structured in three parts: (1) generalized linear mixed models, (2) measurement models and (3) structural equation models. In each part, we start with the simplest version of the model, motivating each extension through examples. Methods of estimation and prediction will also be surveyed. We then consider real applications, specifying models to address the research question, interpreting parameter estimates and providing further insight and model diagnostics using graphical displays of both data and model predictions. We will use consistent notation throughout, emphasizing the communalities between model-types.
This course will benefit statisticians and graduate students in statistics familiar with generalized linear models.
|
| |
