APPLIED MATHEMATICS & STATISTICS (AMS)
Bayesian hierarchical/multilevel and
latent-variable (random-effects) modeling
visits since creation on 30 May 2010.
Venue: one-day short course
sponsored by the School of Mathematics, Statistics and Applied
Mathematics at the National University of Ireland, Galway (NUI
Galway), BIO-SI and the Science Foundation of Ireland
Date: Tuesday 1 June 2010 (11am-5pm).
Location: Information Technology Building, room 202, NUI
Additional details on the course may be found here and here.
Lecture notes and articles: (PDF
Part 1 (Formulation of Bayesian models and fitting them with MCMC in
WinBUGS; case study: measurement of physical constants): a Bayesian overview and lecture notes: part 1
Lecture notes: part 2 (Fixed- and
random-effects models in meta-analysis; case studies: effects of aspirin on
mortality for heart attack patients, effects of teacher expectancy on
student performance), and a third meta-analysis example: Adding art to the rigor of statistical science
(Leonhardt D (2001), New York Times)
Lecture notes: part 3 (Fixed- and
random-effects additive, multiplicative and hierarchical regression models
for count data; case study: effects of in-home geriatric assessment on
hospitalization rates for elderly people living in the community)
Part 4 (Random-effects logistic regression models for multilevel data sets
with binary outcomes; case study: effects of socio-economic and other
variables on type of prenatal care for Guatemalan women): A comparison of Bayesian and likelihood-based methods for
fitting multilevel models (with discussion) (Browne W, Draper
D (2006), Bayesian Analysis, 1, 473-550)
Part 5 (Bayesian hierarchical modeling as an alternative to variable
selection in generalized linear models; case study: prediction of body fat
in humans): Bayesian model averaging: a tutorial (Hoeting JA,
Madigan D, Raftery AE, Volinsky CT (1999), Statistical Science,
14, 382-417) (with comments by Merlise Clyde, David Draper, Ed
George, and a rejoinder by the authors)
Data files and code: (text
R code to do Gibbs sampling
in the aspirin meta-analysis case study, and WinBUGS model, data and initial values files for that example.
WinBUGS model, data and initial values files for the
teacher-expectancy case study.
WinBUGS files for the IHGA case study: model 1, data 1, initial values 1, model 2, data 2, initial values 2, model 3, initial values 3, model 4, initial values 4.
WinBUGS files for the NB10 case study: model 1, data, initial values 1, model 2, initial values 2.
(last modified: 30 May 2010)