Workshop on mixed models in Stata
Thursday 21 September, 2006
Monash University Conference Centre, Collins St, Melbourne
Presenters: StataCorp led by Bobby Gutierrez
A workshop delivered by StataCorp aimed at people who are already familiar with Stata.
The day after the 2nd ANZSUG there will be a 1-day workshop on mixed models in Stata. This will be a hands-on opportunity to learn about mixed models (often called random effect models) and their implementation in Stata. The application of these models includes variance components analysis, longitudinal (panel) data analysis, and hierarchical (multilevel) regression. The workshop will focus on continuous-scale dependent variables but will also discuss categorical dependent variables.
Bring your own laptop if possible although a small number of laptops are available for earlybirds. Material to be covered will include multilevel and longitudinal modelling using Stata with an emphasis on -xtmixed-, but also some -gllamm-.
Program for Thursday September 21st 2006
Analysis of Mixed Models using Stata
------------------------------------
Abstract: This course is a survey of mixed-model analysis using Stata,
focusing on the use of -xtmixed- (released with Stata 9) for linear mixed
models and on the use of -gllamm- (Rabe-Hesketh et al.) for mixed models with
non-Gaussian responses. A prior familiarity with basic Stata is assumed, but
knowledge of mixed models (theory and applications) is not. Throughout, the
lectures will be centered around case studies using real data.
| 9:00 - 10:30 |
Session 1: Theory and Terminology; Topics to be covered include basic model statement, Laird and Ware (1982) panel representation, extension to multiple levels, ML vs REML
estimation, history/connection to classic ANOVA, terminology across various scientific fields, basic -xtmixed- syntax and output |
10:30 - 10:50 |
Morning coffee |
| 10:50 - 12:30 |
Session 2: Using -xtmixed- Part I; Topics to be covered include random-intercept models, random-coefficient models, analysis of growth curves, alternate correlation structures, LR tests
of nested models, best unbiased linear predictions (BLUPs) & empirical Bayes estimates of random effects. |
| 12:30 - 1:30 |
Lunch |
1:30 - 3:10 |
Session 3: Using -xtmixed- Part II; Topics to be covered include advanced model specifications, exploiting block-diagonal covariance structures and -xtmixed-'s unique factor notation, crossed-effects models & tricks for computational efficiency. |
| 3:10 - 3:30 |
Afternoon tea |
| 3:30 - 5:00 |
Session 4: Generalized Mixed Models using -gllamm-; Topics to be covered include binary responses, count responses, ordinal responses, basic -gllamm- syntax & diagnostics using -gllapred-. |
|