36-905: Seminar on Latent Variable Models

4th Mini, 2009
M, Th 3:00-4:20 
Location BH232Q

Brian Junker, Statistics
Richard Scheines, Philosophy

In this mini-semester reading seminar we will read and discuss papers
on latent variable modeling in statistics, especially as they relate
to generative and causal modeling in psychology, the social sciences,
and machine learning.  The course will be lead by Brian Junker, with
interests in latent variable models for psychological testing and
other behavioral data, and Richard Scheines, with interests in latent
variable models and scientific reasoning in the social sciences.  We
will also explore use of latent structures in generative models.

Likely areas for reading include the following.  Some readings will be
historical but we will try to concentrate on more modern treatments.

* Latent variables and individual differences in psychology

* Can you tell from the data that a latent variable *must* exist?

* Identifiability and estimation problems in latent variable modeling

* Latent variables and causal explanations

* Latent structures in generative models

After some introductory lectures, alll papers will be divided among seminar
participants for presentation each week.