ST740 -- Bayesian Inference
Fall 2009 --- 3 credits

---

Instructor: Brian Reich (reich@stat.ncsu.edu)
Office hours: 4262 SAS Hall, MW 1-2pm
Course times/location: MWF 10:15-11:05, 5270 SAS Hall
Syllabus: http://www4.stat.ncsu.edu/~reich/st740/Syllabus

Course assignments

---

• HW 1 (Due: Monday, 8/31): 1.6, 1.8, 2.2, 2.3, 2.10
  
• HW 2 (Due: Friday, 9/11): 2.7, 2.9, Appendix B problems 1, 3, and 7
  
• HW 3 (Due: Wednesday, 9/23): 3.5a, 3.6, 3.7, 3.9a
  
• HW 4 (Due: Monday, 10/12): 3.9b, 3.16bc, 3.20
  
• HW 5 (Due: Monday, 11/4): PDF (R code, BUGS code, 2008 election results, and NC county adjaceny matrix).
• HW 6 (Due: Monday, 11/23): 2.18, conduct a brief simulation study to show that modeling the residual distribuiton in a linear regression using a DP can improve estimation of the regression coefficients.
  
  
  
• Data for the homework assignments
  
  
• Midterm 1 (In class 9/25):
  
  
• Midterm 2 (In class 11/6):
  
  
• Final project (Proposal due 11/9; Report due 11/30/1; In-class presentations 11/30/1-12/16):

Course outline and important dates

---

Week -- Description
8/19   -- Introduction
8/24   -- Bayes Basics
8/31   -- Prior distributions; Empirical Bayes
9/7     -- Labor Day (M); Bayesian performance
9/14   -- Bayesian performance
9/21   -- MCMC sampling; Midterm Exam (F, 9/25)
9/28   -- MCMC sampling
10/5   -- MCMC diagnostics; Fall Break (F)
10/12 -- Linear/Generalized linear models
10/19 -- Hierarchical models
10/26 -- Model selection and diagnostics
11/2   -- Model selection and diagnostics; Midterm Exam (F, 11/6)
11/9   -- Nonparametric Bayes
11/16 -- Case Studies
11/23 -- Case Studies; Thanksgiving (WF)
11/30 -- Student Presentations
12/16 -- Final exam, Wednesday, 8-11am, 5270 SAS Hall

Computing/data/references

---

• R code to the plot the prior/posterior for the beta/binomial model
• Hiance A, Chevret S, Levy V (2009). A practical approach for eliciting expert prior beliefs about cancer survival in phase III randomized trial. Journal of Clinical Epidemiology, 62, 431-437.
• Gelman A (2006). Prior distributions for variance parameters in hierarchical models. Bayesian Analysis , 1, 515-533.
• WinBUGS code for Poisson regression.
• WinBUGS code for Poisson regression.
• R code to the plot the bias, variance, and MSE for the beta/binomial model
• Simulation code in R to illustrate the Stein effect
• R code to compare interval estimates for the binomial proportion
• Gelman A (2008). Objections to Bayesian statistics. Bayesian Analysis , 3, 445-450.
• Introduction to R slides (Cheat sheet)
• R code to perform univariate rejection sampling
• MCMC code for the Bayesian t-test
• MCMC code for the Bayesian linear model
• MH code for the Bayesian logistic regression model
• Sample WinBUGS code.
• R code to call WinBUGS to analyze this model.
• Gumbel WinBUGS code.
• Charlie Geyer's thoughts on MCMC diagnostics including multiple chains and burn-in.
• WinBUGS code for a linear model with missing data.
• WinBUGS code for robust linear model.
• WinBUGS code for longitudinal data.
• R code to compute DIC for the one-way random effects model.
• WinBUGS code for a stochastic search variable selection.
• Reich BJ, Storlie CS, Bondell HD (2009). Variable selection in Bayesian smoothing spline ANOVA models: Application to deterministic computer codes. Technometrics, 51, 110-112.
• Semiparametric regression for the motorcycle data.
• R code to make draws from a DP/DPM prior.
• R code for linear regression with DP mixture model for the residuals.
• Berry DA (2006). Bayesian clinical trials. Nature Reviews, 5, 27-36.
• Bayesian sample size calculations.