Methods and Statistics 2: Structural equation modelling and multilevel analysis

Course goals

Content

The course is organized in 2 parts in a flexible format of 5 weeks each.

Part 1: Structural equation models (5 weeks).
The SEM framework integrates two types of models. First are simultaneous equations, i.e., an integrated set of regression models allowing the dependent variable (a `consequence') of one equation to be an independent variable (a `cause') in another equation. Second, `factor analytical' models for the reflexive measurement of latent variables will be introduced. These are widely applied for measuring attitudes, intentions, etc. Topics to be discussed: model specification; model fit; decomposition of effects; categorical variables; multiple group analysis; testing for structural and measurement invariance with dependent and independent groups; interactions involving latent variables; SEM for longitudinal data; latent curve models.
Class schedule: twice a week a short lecture, followed by a short computer practical (on te same day)

Part 2: Multi-level models (5 weeks).
These models are often used for the analysis of `hierarchical problems' in which the causes of outcomes (e.g., the performance of pupils in schools) are located at the level of the individual (e.g., own and parental resources) as well as in the context shared by some of the individuals (e.g., characteristics of the class and of the teacher). Data with this structure violate the assumption of `independent observations' made in, e.g., standard regression analysis. Multilevel models can also be used with longitudinal designs (`time points within persons', panel data). We will focus mostly on the versions of the model with a `continuous' response variable, Time permitting the modification to the binary case, designs with more than two levels (e.g., students within classes within schools), cross-classified designs, and multilevel extensions of SEMs (multilevel path models, multilevel factor analysis) will be introduced.
Class schedule: each week, one 3h lecture and one 3h computer practical on different days.