Structural Equation models (SEM) are traditionally linear models in which latent variables (factors) play a dominant role. In general there is a distinction between two parts in SEM: the measurement part and the structural part of the model. In the measurement part of the model the question is whether there is a factor underlying a subset of the observed variables. The structural part of the model is mainly concerned with the relationship between the factors. Extensions of the traditional structural equation models are: categorical latent (and observed) variables, i.e , Latent Class Analysis (LCA), Item Response models (IRT), growth models, Multi-Level (ML) models, and Mixture Models for dealing, e.g., with non-normally distributed variables. A more common name for this class of models is nowadays Latent Variable Models (LVM).
In the second part of the course we will study graphical models. Here graph theory is used to describe and analyze causal models. Structural equation models form a special case. Graph theory allows us to visualize these models and also provides new computational tools for estimation and testing.
The course consists of 7 lectures and 7 assisted working groups
There will be 4 practical assignments, that are graded.