Contact InformationCatalog DescriptionWhat This Course Is Really AboutCourse RequirementsTextbooksTentative Course PlanIntroductionCausal Inference and Research DesignDescriptive StatisticsCorrelation and RegressionMidterm Exam and Student PresentationsProbabilitySamplingStatistical InferenceFinal Exam and Student PresentationsPOL 451: Statistical Methods inPolitical ScienceFall 2007Kosuke ImaiDepartment of Politics, Princeton University1 Contact InformationOffice: Corwin Hall 041Phone: 609–258–6601Fax: 973–556–1929Email: [email protected]: http://imai.princeton.eduOffice Hours: Mondays 3:00pm – 4:00pm or by appointment2 Catalog DescriptionMondays and Wednesdays, 1:30pm – 2:50pm.Friend Center 202.In this course, students will learn basic research design and data analysis methodology inempirical social science research. The main goal is to learn how statistical theory can be usedto make causal inferences in experimental and observational studies. The course satisfies theanalytical methods requirement for politics majors. The materials of this course are particularlyuseful for those who plan to use quantitative analysis in their junior papers and senior thesis aswell as for those who wish to apply for graduate programs in the social sciences. Familiarity withelementary probability theory is helpful, but is not required.3 What This Course Is Really AboutThis course introduces the basics of statistics and probability to students with little or no mathe-matical background. It is intended as a first course on probability and statistical methods in thesocial sciences. One main focus of the course is the key question of how to use statistics to makecausal inferences, which are the main goals of most social science research. Descriptive inferencesand survey sample surveys are also covered. Students will learn basic data analysis techniques aswell as elementary statistical and probability concepts.14 Course RequirementsThe final course grade is based on in-class participation (20%), problem sets (30%), and the choiceof a project or exams (50%).• In-class participation: This is a seminar course, and regular attendance is required. Stu-dents are expected to read the assigned materials before classes and actively participate indiscussion.• Problem sets: There will be several problem sets during the semester. Students are al-lowed to discuss the questions with each other, but they must write up their answers to thequestions on their own. Late submission will be penalized (for each day, 30% reduction willbe applied).• Research project: If you choose this option, your research project should contain anoriginal analysis of a data set addressing substantive questions of interest. Some potentialprojects are:– analysis of a data set that you collected, an experiment that you conducted, or a surveythat you administered.– the empirical analysis portion of your senior thesis or junior paper.– your reanalysis of a data set that has been analyzed in a published scholarly paper.By the end of the semester, you must complete a research report that describes a substantivehypothesis of interest and explains the methods and results of the empirical analysis. Thisreport is due at 5pm on January 21, in my departmental mailbox (Corwin Hall). Latesubmission will not be accepted.To support continued progress for the project, students are required to:1. meet with me to discuss their project during September.2. give a 15 minute in-class presentation about research design right after the fall break.3. give 15 minute in-class presentation about preliminary results in December.The presentations will be 20% of the c ourse grade, whereas the final project report will be30%. In addition, I will be available for individual meetings anytime during the semester todiscuss projec ts and comment on drafts.• Exams: If you choose this option, you are required to take a mid-term exam (October24) and a final exam (the date specified by the registrar). Both exams will be in-class andclosed-book. The mid-term exam will be 20% of the course grade, and the final exam willbe 30%.5 TextbooksThe main textbook we use in this class is:• Freedman, David, Robert Pisani, and Roger Purve s. (2007). Statistics 4th eds. Norton.2This book is available for purchase at the U-store and is also reserved at the library. The coursealso relies on journal articles and book chapters.In addition to the required textbook, you may find some materials from the following textbooksuseful.• Achen, Christopher. (1982). Interpreting and Using Regression Sage.• Fox, John. (2002) An R and S-PLUS Companion to Applied Regression Sage.• DeGroot, Morris H. and Mark J. Schervish (2002) Probability and Statistics, 3rd Edition,Addison Wesley.The first is an introductory textbook to regression, and the se cond one is a reference book forthe statistical software R, which will be the (free!) software for data analysis in this course; thisbook is available for purchase at U-store. The third book is a more advanced (i.e., mathematical)textbook on probability and statistics.6 Tentative Course Plan6.1 Introduction1. (September 17). Overview(a) What is statistical inference?(b) When are statistics useful?• Salsburg, David. (2002). The Lady Tasting Tea: How Statistics Revolutionized Sciencein the Twentieth Century Owl Books.• Taleb, Nassim Nicholas. (2007). The Black Swan: The Impact of the Highly ImprobableRandom House.6.2 Causal Inference and Research Design1. (September 19). Randomized Experiments vs. Observational Studies(a) What is causal inference?(b) Statistics and causal inference(c) Advantages and disadvantages of randomized experiments and observational studies• Freedman et al. Chapters 1 and 2.• Holland, P. W. (1986) “Statistics and causal inference” (with discussions) Journal ofthe American Statistical Association 81, 945–970.• Taubes, Gary. (2007). “Unhealthy Science” The New York Times Magazine September16, Section 6.2. (September 24 and 26). Experiments in the Real World3(a) Pros and cons of randomized experiments(b) Applications of field, social, and natural experiments• Burtless, Gary. (1995). “The Case for Randomized Field Trials in Economic and PolicyResearch.” Journal of Economic Perspectives 9, 63–84.• Heckman, James J. and Smith, Jeffrey A. (1995). “Assessing the Case for So cial Ex-periments.” Journal of Economic Perspectives 9, 85–110.• Bertrand, Marianne and Mullainathan, Sendhil.