Bayesian Statistics, 22S:138Fall 20091 General InformationInstructor: Kate Cowles, 374 SH, [email protected] hours: T 1:30 - 2:20 p.m.W 12:30 - 1:20 p.m.Th 1:30 - 2:20 p.m.Please feel free to make appointments to see me outside of office hours,and to send me questions by e-mail.Department: Statistics and Actuarial Science, 241 SHDEO: Luke Tierney, 241 SH, 335-0712Lectures: M, W, F 11:30 - 12:20 61 SHLab: Alternate F 11:30 - 12:20 41 SHWeb page: http://www.stat.uiowa.edu/~kcowles/s138_2009Handouts, homework assignments, datasets, etc.will be posted on the web page for you to download.Textbook: Gelman, Carlin, Stern, and Rubin, Bayesian Data Analysis2nd ed., Chapman & Hall/CRC, 2004On reserve: Berry, Statistics: A Bayesian PerspectiveMath Sciences Library Johnson and Albert, Ordinal Data Modeling125 McLean Hall Lee, Bayesian Statistics: An IntroductionCarlin and Louis, Bayes and Empirica l Bayes Methods for Data AnalysisWinBUGS manuals2 Course goals and objectivesThrough hands-on experience with real data from a variety of applications, students willlearn the basics of designing and carrying out Bayesian analyses, and interpreting and com-municating the results. Students will learn to use software packages including WinBUGSand BOA to fit Bayesian models.3 Evaluation of students3.1 HomeworkHomework assignments will consist of data analysis on the computer, written interpretationof computer output, and other written questions. In general, homework will be assignedeach Fri. and will be due in class the following Fri. Exceptions to this schedule will beannounced in class.1Show your work when solving written homework problems. For computer problems, turnin printouts of your commands or programs and their output.You are encouraged to study with others. However, if you do work with others on home-work assignments, please: a) write up your own ass ignment and make sure you completelyunderstand all solutions that you submit, and b) write the names of the others in yourstudy group on your assignment.Late homework will not be accepted except as required by university policy, i.e. because of“illness, mandatory religious obligations, or other unavoidable circumstances or Universityactivities.”3.2 ProjectsStudents will work in groups of three to carry out projects involving application of Bayes ianmethods to problems of their own choosing. Some examples are:• Carry out a complete Bayesian analysis of a real dataset. This might involve:– description of the research question and dataset– specifying an appropriate Bayesian model– determining appropriate values for prior parameters– fitting the model using WinBUGS– checking convergence– analyzing the output using BOA– reporting and interpreting the results• Compare different me thods of fitting the same model to the same dataset– normal approximations– MCMC– other simulation methods– analytical computation (if feasible)– etc.• Carry out a Bayesian analysis of a dataset for which a classical analysis has beenreported in a journal. Compare and contrast the results obtained by the two ap-proaches.• Fit a Baye sian model to a dataset using several different choices of prior (hyperpa-rameters and/or functional form). Discuss the meaning of the different results, andthe robustness of the model to prior specifications.• Fit several different plausible Bayesian models to the same dataset. Carry out a checkof model adequacy and model fit. Discuss the results.2• There are endless other possibilities. Find something that interests you,or see me forideas.I w ill expect more sophisticated projects from graduate students.Projects will be carried out in three phases. Please meet with me at least once while youare working on each phase.• Project proposal (due 11/02)This is a detailed description of what you plan to do, including question(s) to beaddressed, dataset to be used, methods to be applied. Also specify the method ofpresentation that you intend for the final project. (See below.)• Project interim report (due 11/18)This informal report will indicate that your project is “on track.” All computingshould be done at this time. The report will include results obtained thus far anda brief summary (hand-written is O.K.) of what they mean and what remains to bedone. In addition, the report will include a list of the tasks performed by each memberof the project team.• Project presentation (papers or presentation materials must be posted or submittedby 12/07)Projects must be finalized in a form that can be shared with the entire class, such as:– posting a document on the course web page– preparing a poster– giving an oral presentation with overheads, slides, or computer imagesPosters and oral presentations will be given in class during the final week of classes.3.3 ExamsThere will be two 1-hour midterm exams and one comprehensive 2-hour final. Studentsmay bring one 8-1/2 x 11 in. sheet of paper with notes to each midterm, and 3 sheets tothe final.Missed exams may be made up only with documentation of reasons required by universitypolicy (see “Late Homework” above).Exam dates and times:Midterm 1 Fri. 10/02 in classMidterm 2 Fri. 10/30 in classFinal Mon. 12/14 2:1533.4 GradingThe course c omponents will be weighted as follows:Homework 10%Midterms 35% (17.5% each)Project 20%Final 35%Grading will be on a curve, with +/− grades used. A grade of A+ represents exceptionalwork and rarely is awarded.4 Extra HelpThe Statistics Tutorial Lab, located in 202 CC, gives free tutorial assistance to students in22S:2, 8, 25, and 39. In addition, several graduate students have volunteered to indepen-dently tutor students in various 22S: courses at mutually- arranged times and fees. Pleasecheck the web site www.stat.uiowa.edu/courses/tutoring.html for tutoring details.5 Campus Wide Flu Like Illness Classroom Absence PolicyPublic health authorities have recommended that people with flu-like illnesses stay homeand not return to public space s until 24 hours after they have no fever. In order to preventthe spread of disease, please do not come to class, meet with other groups of students, attendoffice hours, or contact offices in person while you are ill. Base d on this recommendation, Iwill not require you to report to a doctor or to Student Health to verify a flu-like illness. Ifyou are ill, please complete an illness-absence form (http://www.registrar.uiowa.edu/forms/H1N1_absence_form.pdf ) when you are
View Full Document
Unlocking...