2011/12 – Mixed models for multi-level and panel data

Název kurzu: Mixed models for multi-level and panel data

Modul: kvantitativní

Kód: PSY028d a SOC931d

Vyučující: Prof. dr. Herman van de Werfhorst

Termín konání: 6. 11. února 2012

Vyučovací jazyk: angličtina

Rozvrh:

po 6.2.: 9-12, 13-16 PC 26

út 7.2.: 9-12,18-20 PC 26

st  8.2.: 9-12, 13-16 PC 25

čt 9.2.: 9-12,13-16 PC 25

pá 10.2.: 9-12, 13-16 PC 25

Outline

ANNOTATION – BRIEF DESCRIPTION OF THE METHOD:

In this course, students will be trained in the understanding and application of mixed models. Mixed models are characterized by a combination of fixed and random effects.

A well-known application of mixed models are better known as multi-level, or hierarchical regression models. Multilevel models are adequate to deal with nested data, for instance when individuals are nested within school classes, or citizens are nested in countries. Also more than two levels can exist. Multilevel models allow for the inclusion of explanatory variables at the individual and contextual level(s) (e.g. student characteristics like gender and characteristics of the teacher shared by all students within a school class).

BRIEF DESCRIPTION OF THE COURSE :

During the course, first the logic of fixed effects models is explained. Then an extension will be given to adding random parts to the model. First attention is given to variance components, so students learn which proportion of the total variance in the dependent variable can be attributed to the different levels. Then, models will become more complex, by adding explanatory variables at all levels, and adding random slopes. Special attention will be given to interaction effects between variables at different levels (cross-level interactions). Most of the applications will be made using data from the Programme for International Student Assessment 2006 data (PISA 2006), and/or the European Social Survey (ESS). Students are also encouraged to bring their own data to the course. Most applications will be done using the xt packages in Stata, although some introduction will also be given to the Stata package Gllamm.

SOFTWARE: Stata version 10 or 11.