MAS655 Graphic Models in Statistics
Spring 2010

Instructor: Sung-Ho Kim
           Office: E6-1 Rm3411               Phone: 042-350-2737
           Email: sung-ho.kim at kaist dot edu      Office hours: TTh 14:00-15:00

Lecture hours: TTh 09:00-10:15

Text:   (1) Graphical Models by S.L. Lauritzen
           (2) Graphical Models in Applied Multivariate Statistics by J. Whittaker
           (3) Lecture notes

Course Topics:
1. Graphs and hypergraphs
2. Conditional independence and Markov properties
3. Hierarchical models and decomposable models
4. Chain graph models
5. Collapsibility
6. Graphical modeling

Grading:  Homeworks: 50 % or more,  Exam: 50% or less.

Note:     
This course may not be appropriate for the students who are not familiar with linear models and statistics.  Homework problems will be given in class.

Course Schedule:

week    contents
1           Independence and conditional independence
2           Graphs and decomposable graphs
3           Hypergraphs and forests
4           Conditional independence and Markov properties
5           Contingency tables and log-linear models
6           Estimation in hierarchical log-affine models
7           Collapsibility
8           Mid-term exam
9           Multivariate normal models
10         Covariance selection models and Gaussian decomposable models
11         Maximum likelihood(ML) estimation for covariance selection models
12         Conditional Gaussian model
13         Graphical interaction model for mixed data
14         ML estimation for decomposable models with mixed data
15         Chain graph models for mixed data
16         Final exam or term project