I. Course DescriptionII. Course PrerequisitesIII. Course Goals and ObjectivesIV. Methods of Instruction and Work ExpectationsV. Course Text and ReadingsVI. Course Outline/Weekly ScheduleVII. Evaluation and GradingVIII. Other Course Information and PoliciesCourse SyllabusPubH 7440 Introduction to Bayesian AnalysisSpring 2012Credits: 3Meeting Days: Tuesday-ThursdayMeeting Time: 2:30-3:45pmMeeting Place: Mayo C381 (SPH Computer Lab)Instructor: Prof. Brad CarlinTeaching Asst: Harrison QuickOffice Address: A460 Mayo Bldg, MMC 303, 420 Delaware St S.E., Minneapolis MN 55455Office Phone: (612) 624-6646Fax: (612) 626-0660E-mails: [email protected]; [email protected] blog: http://blog.lib.umn.edu/quic0038/pubh7440/Office Hours: Brad: TuTh 3:45-5:15 pm or by appointment Harrison: Tues 11:00am-1:00pm and Thurs 12:00-2:00pm; held in TA Room, Mayo A446I. Course DescriptionThis course introduces hierarchical Bayesian statistical methods that enable investigators to combine information from similar experiments, account for complex spatial, temporal, and other correlations, and also incorporate prior information or expert knowledge (when available) into a statistical analysis. The course explains the theory behind Bayesian methods and their practical implementation, and also compares them with classical (frequentist) methods. The course emphasizes data analysis via modern computer methods viathe WinBUGS and R freeware packages that are introduced and used throughout the course. II. Course PrerequisitesStat 5101-02 or PubH 7405-7406 or instructor's consent. If you are unsure about your qualifications for the course, please contact the instructor.III. Course Goals and ObjectivesUpon successful completion of the course, students will be able to independently formulate Bayesian hierarchical models for analyzing complex datasets arising from non-trivial statistical designs and observational data settings. They will also be able to implement these models using statistical software, and write and give comprehensive oral reports of their analyses. IV. Methods of Instruction and Work ExpectationsMethods of instruction will be through in-class lectures and presentations, and also through hands-on practice with the WinBUGS and R software in the SPH Computer Lab (Mayo C381).1V. Course Text and Readings The only required text for the course is Bayesian Methods for Data Analysis, 3rd edition, by Bradley P. Carlin and Thomas A. Louis. Other readings/websites will be provided as needed. VI. Course Outline/Weekly Schedule Week 1 (1/17): Preliminaries; overview and basics of Bayesian inference Week 2 (1/24): Introduction to the R computing environment and language; basics of Bayesian computing Week 3 (1/31): Theory of Bayesian linear models; Bayesian linear models in R Week 4 (2/7): Introduction to WinBUGS and hierarchical modeling Week 5 (2/14): Bayesian computing; Markov chain Monte Carlo (MCMC) methods; packages in R Week 6 (2/21): Bayesian model criticism and selection Week 7 (2/28): Review; MIDTERM 1 (in-class) on Thursday Week 8 (3/6): Empirical Bayes methods: point and interval estimates, frequentist comparisons Week 9 (3/20): Bayesian design and analysis of clinical trials Week 10 (3/27): Hierarchical longitudinal and time-series models; MIDTERM 2 (take-home) assigned Tues Week 11 (4/3): Bayesian survival analysis and frailty modeling Week 12 (4/10): Review; MIDTERM 2 DUE Tues 11:55 pm; Project Selection DEADLINE Thurs 5:00 pm Week 13 (4/17): Spatial and spatiotemporal models Week 14 (4/24): Bayesian Case Studies; Special Topics; Review and Catch Up Week 15 (5/1): FINAL PROJECT PRESENTATIONS (may carry over into finals week, depending on class size); ---------------------------------- Final Project writeups DUE Thurs May 10 5:00 pm) -------------------------------------------VII. Evaluation and GradingYour final grade will be based upon homework assignments (35%), two midterms (15 and 20%, respectively),and a final project (30%). The homework problems will include theoretical and applied questions, mostly from the text. Assignments will be given out as appropriate throughout the semester, and will generally be due one week after they are assigned. Students should try to do their own work on these problems; the TA and I are available for questions, of course. The first midterm will be in-class (open-book), while the second will be take-home. For data analysis homework and midterm problems, your write-up must be a careful reportof your models, methods, interpretations, and conclusions -- as if you were making a final report to a supervisor who has statistical training, but doesn't want to get bogged down in the details. Include only the relevant parts of your computer output in your report, labeling all plots, variables, and so forth. You need not get too carried away -- always substitute prose for output where possible. The final project involves preparing a short (5-10 page) paper and giving a brief (15-20 minute) classroom presentation on some subtopic of interest to you. This will be a group project; students must form into groups of 2 or 3 persons. (While I realize you many of you may prefer to work on your own, part of the experience here is forcing you to have to work and negotiate with someone else.) Group members may assign speaking responsibilities in any way they please, but writing responsibilities should be shared. All members of a group will receive the same final project grade. Once your group has formed and identified a topic of interest, you will need to meet briefly with me to “reserve” your topic. I may suggest a paper or two for you to read, which may in turn suggest several interesting project possibilities: extending an analytical result, simulating the performance of some procedure, undertaking a challenging data analysis, etc. More final project information will be provided as the course unfolds. I take a very dim view of unexcused late assignments, especially in a class like this where most of the work is``take-home.'' As a general rule, prior notification is essential to my accepting a late paper of any kind. If illness or travel is going to cause you to miss a deadline, don't surprise me -- call or send an e-mail message (as crazed modern academics, the TA and I check our voice messages and e-mails constantly).2Incomplete GradeA grade of incomplete “I” shall be assigned at the