Bayesian Reading Group

Meeting Times and Logistics

bullet FRIDAYS, 11-12, Economics Conference Room, SSH1131
bullet First Meeting: October 14

Textbook

Koop, Gary (2003) Bayesian Econometrics, Wiley. The book is not currently available at the bookstore nor at the main library (it is available in the ARE Library). It is probably easiest to purchase it from Amazon.

The website for the book contains all the data sets and MATLAB code.

 

Schedule of Tasks

bullet October 21 meeting: read chapter 1 on your own and try to do the exercises. We will discuss chapter 2, which introduces linear regression from a Bayesian perspective.
bullet October 28 meeting: by now you should have read chapters 1-3. It would not hurt to have read chapter 4. We will emphasize computational Monte Carlo techniques. Check the handouts and examples we have prepared.
bullet November 4 meeting: we will finish discussing computational MCMC techniques. These are scattered in chapters 3, 4, and 5.
bullet November 18 meeting: you should read chapter 9. This is a jump in the order of the chapters but a natural extension of the linear regression model with latent variables (e.g. probit). We will see how bayesian methods can be used for data augmentation. Also, please note Colin has prepared notes and Stata code for the probit example. Please check the handouts section.
bullet November 25 meeting: Thanksgiving!
bullet December 9 meeting: Tim Cogley will discuss Bayesian techniques for time series models. Tim has experience estimating time-varying parameter VARs and other models. In preparation for that meeting, Tim has asked that you read the following paper: "Has Inflation Become Harder to Forecast" by Stock and Watson, JMCB (2005)

 

Handouts

bullet Chapter 1: Brief notes and exercises. Exercise solutions.
bullet Chapter 2: Notes on the role of the prior; other updating examples (by Jim Chalfant)
bullet Chapter 3: Introduction to Bayesian Computation; Gibbs sampler example 1 (in GAUSS); Gibbs sampler example 1 (in STATA) and sample output (STATA)
bullet Chapters 1-4: Summary of important concepts so far
bullet Chapters 1-9: Revised and updated summary of important concepts. Stata code for Probit example with output file.

 

Other Campus Events

bullet October 28 at 2pm in IGA: "A Pragmatic Justification for the Use of Bayesian Methods in the Social
Sciences" by Andrew Martin, Washington University of St. Louis. Click here to download his slides.
 

Software

According to Koop, most Bayesian econometricians use MATLAB, which are available both in Econ and ARE. MATLAB is a matrix language, similar to GAUSS. GAUSS users should have low barriers in making the switch (in fact there are some gauss-to-matlab converters on the web). Here is a list of the leading resources available for free:

bullet Bayesian Analysis, Computation and Communication (BACC): This website contains free Bayesian routines that work in Matlab, S-Plus, and R. There are some older version that will work with GAUSS as well, available upon request. The software is part of an NSF project led by Siddharta Chib and John Geweke, two leading expert in Bayesian Econometrics.
bullet Bayesian Inference Using the Gibbs Sampler (BUGS): This is a stand alone program, the latest version of which has a windows interface. There is a short movie available that shows how this interface works. Akin to Eviews in terms of having canned procedures at the ready, probably less useful if you want to estimate a specific model not pre-canned.

Other Sources of Software

bullet Gary Koop's MATLAB code
bullet James LeSage's MATLAB routines (includes other classical routines)
bullet Andrew Martin's MCMCpack (works with R)

 

Useful References

bullet An Introduction to Modern Bayesian Econometrics by Tony Lancaster (Blackwell). This is another good introductory textbook. The book subscribes more strongly to the Bayesian point of view. There are numerous exercises and examples in winBUGS (the windows version of BUGS).
bullet Bayesian Data Analysis by Gelman, Carlin, Stern, and Rubin (Chapman and Hall). This is a higher level book from a more statistical perspective. The website for the book contains the solutions to exercises and examples of programs in BUGS.
bullet Using Simulation Methods for Bayesian Econometric Methods: Inference, Development and Communication by John Geweke, Econometric Reviews (1999)

 

Other Useful Links

bullet Jim Hamilton's (UCSD) introductory class notes
bullet Andrew Martin's (WUSTL): website; Monte Carlo movie; Gibbs sampler movie; Metropolis-Hastings movie