Business 41914: Multivariate Time Series Analysis

Spring Quarter of 2007

Instructor: Ruey S. Tsay

Office: HPC 455
Tel:      773-702-6750
Fax:     773-702-0458
e-mail: ruey.tsay@ChicagoGSB.edu

Lecture: Wednesdays 8:30 am to 11:30 am, HPC 04

Office hour: (a) Wednesdays 1:30 pm to 2:30 pm
                    (b) By appointment
You may e-mail me questions. E-mail is the easiest way
to make contact with me. I try to check the e-mail at least
once a day.

Teaching Assistant:
Mr. David Matteson, his e-mail: matteson@uchicago.edu

Text:  No textbook is used

Some reference books:
(a) Time Series Analysis: Forecasting and Control. Box, Jenkins and Reinsel (1994)
      3rd Ed. Prentice Hall.  Chapters 10 and 11.
(b) A Course in Time Series Analysis: Pena, Tiao and Tsay (2001) Wiley
      Chapters 14 and 15.
(c) Time Series Analysis by State Space Methods: Durbin and Koopman (2001)
      Oxford University Press, for Kalman Filtering and Smoothing.
(d) Time Series Analysis: Hamilton (1994) Princeton University Press.
      Chapters 11, 18,  19 & 20.
(e) Analysis of Financial Time Series, 2nd Ed., Tsay (2005), Wiley.
      Chapters 8, 9, 10, & 11.
(f) Additional reference books given in the syllabus.

Some reference articles:
1. Tiao, G. C. and Tsay, R. S. (1989). Model specification in multivariate time series (with discussion). Journal of Royal Statistical Society, Series B, 51, 157-213.
2. Tsay, R. S. (1991). Two canonical forms for vector ARMA processes. Statistica Sinica, 1, 247-269.

Course Syllabus.

Grading: Midterm 40% + Final project 40% + Homework 20%
where scores of each component are normalized to be out of 100.

Final project:  An empirical project or a research article

Midterm: May 9 (Lecture: 8:30-9:30 am, Exam 9:30-11:30 am)
Open books and notes. A calculator is needed.
Exam and Solutions

Lecture:
Week1:  Transfer function model or distributed-lag model.  Chapters 10 & 11
                of Box, Jenkins and Reinsel (1994).
                Data set used: gas-furnace
                Handout: lec1
Week2:  Vector ARMA Models. Chapter 14 of Pena, Tiao and Tsay (2002)
                Data sets used: clsma1.dat  clsar1.dat  m-decile1510.txt
                Handout: lec2   A simple R function to compute multivariate Ljung-Box: mq
Week 3: Vector ARMA Models (continued).
               Data sets used: gas-furnace (see Lecture 1),  CGK series 
               Handout: lec3
Week 4:  Vector ARMA Models and Unit-Root Nonstationarity
               Handout: lec4
               Data set used: q-gdpun.txt (year, mm, dd, gdp, unemp-rate).
Week 5:  Co-integration
               Handout:  lec5,   Data set used: m-bnd.txt   
Week 6: Structural Specification of VARMA Models
               Handout: lec6,       
Week 8: Empirical structural specification: R-Program--Kronid
               Handout: lec7,    data set: flour.dat  
Week 9: Seasonal vector time series 
               Handout: lec8,    data set: house.dat 
Week 10: State-space models and the Kalman filter 
               Handout: lec9   data set:  aa-3rv.txt (5m, 10m, 20m)

Homework assignment:
 HW1:  due on April 4 (before class). Dataset: series
 HW2: due on April 18. Data sets for Q1, Q2, Q3
 HW3: due May 2. Data sets: m-mortg.txt, m-gs2.txt
 HW4: due May 23. Data sets: (1) m-gs7.txt, m-gs2.txt, (2) hw4a.txt, (3) & (4) hw4b.txt

Solutions to HW Assignments:
  HW1 & output.
  HW2 & output.
  HW3.

Final Project:  Due on Week 10 (Friday)
(a) Individual project or a team of two students
(b) Theoretical project: journal article critique or empirical project: real data
      of vector time series (at least 2-dimensional).
(c)  Students can also use empirical data to test any theorey relevant to the course.

Computing:
SCA is used for most analyses. R and S-plus are also used for
Kalman filter and multivariate volatility modeling. Other packages
can also be used.  We shall give some demonstrations in class.
Students may use other packages that are not discussed in class.

Old in-class exam: exam05 Scaoutput