Andrew Gelman's courses at Columbia

• All courses are (officially or unofficially) cross-listed in statistics and political science.
• Students in statistics and in applied fields (especially, but not limited to, the social sciences) are welcome.

• Research in quantitative political science, Pols/Stat 8991 (co-taught with Shigeo Hirano) (every semester; meets once per week for two hours): Our working group in political science. Weekly speakers from Columbia and outside with discussion from students in the class. Participants in the class also discuss their own research. Current topics of interest include red and blue states, representation of provinces within a country, and political polarization.
  
• Applied regression and multilevel models, Stat 6101 or Stat/Pols 4330 (next offered Fall, 2008): A course in applied statistics. First third of the course covers applied linear regression, logistic regression, and generalized linear models, focusing on practical data analysis and computation. Rest of the course covers multilevel regression. Examples from social sciences and public health. Computation in R and Bugs.
  Prerequisite: a course in multiple regression (for example, Pols 4910/4911, Soci 4074/4075, or Stat 4315)
  Textbook: Data analysis using regression and multilevel (hierarchical) models," by Andrew Gelman and Jennifer Hill
  
• Bayesian data analysis, Stat 6102 (next offered Spring, 2010?)): Stat 6102 is officially titled "Statistical modeling and data analysis II." We cover Bayesian data analysis--modeling, inference, computing, and model checking. Computation in R.
  Prerequisite: calculus and probability theory. Some statistical knowledge would be helpful also.
  Textbook: "Bayesian Data Analysis," second edition, by Andrew Gelman, John Carlin, Hal Stern, and Don Rubin
  
• The teaching of statistics at the university level, Stat 6600 (next offered Fall, 2008?): Theory and practice of teaching statistics using active learning. Participants in the class practice demonstrations, drills, and other techniques for stimulating student learning.
  Prerequisite: none.
  Textbook: "Teaching Statistics: A Bag of Tricks," by Andrew Gelman and Deborah Nolan (Oxford), "First Day to Final Grade: A Graduate Student's Guide to Teaching," by Anne Curzan and Lisa Damour (Michigan), and "Teaching Tips," by Wilbert McKeachie et al. (Houghton Mifflin).
  
• Applied Bayesian statistical computing, Stat 6104 (next offered Spring, 2008): We cover topics in Bayesian computation, statistical graphics, and software validation, as well as special topics that interest the class. There is some homework (writing programs and making graphs in R) and a final project to be done in pairs.
  Prerequisite: none. But if you don't know any statistics, and you don't know any programming, you'll be in trouble.
  Textbooks: "Bayesian Data Analysis," second edition, by Andrew Gelman, John Carlin, Hal Stern, and Don Rubin, and Data analysis using regression and multilevel (hierarchical) models," by Andrew Gelman and Jennifer Hill
  
• Decision analysis, Stat 4419 (next offered 2009/2010?): First quarter of the course covers problem-solving in classical decision analysis--decision trees, expected values, value of information, how to calculate optimal decisions and the value of decision trees. Second quarter covers classical decision theory--derivation of utilities from preferences, utility functions in one and more dimensions, dollar value of a life. Third quarter covers violations of the classical theory--"heuristics and biases" of actual decision making. Fourth quarter of the semester covers applications in social science, public health, and business.
  Prequsisite: calculus, probability theory (for example, Stat 3000, Stat 4105, or the equivalent): understanding of basic probability theory and the willingness to do some integrals (averaging) and derivatives (optimization).
  Textbook: "Making Hard Decisions," second edition, by Bob Clemen (Duxbury Press), and also photocopied readings
  
• Sample surveys, Stat 4335 (next offered 2009/2010?): Theory and applications of survey sampling. Design: simple random sampling, stratified sampling, systematic sampling, cluster sampling, optimal sampling design. Analysis: estimates and standard errors for all designs, ratio estimation, regression estimation, poststratification, survey weighting, missing-data imputation. We refer to applications in social science, public health, and business, but the focus of the course is on learning the fundamental statistical methods.
  Prerequisite: basic statistics (for example, Pols 4910 or Soci 4074), but some regression analysis would be helpful also.
  Textbooks: "Sampling: Design and Analysis," by Sharon Lohr (Duxbury), and "Survey Methodology," by Groves et al. (Wiley)
  

What course should I take?

• Applied regression and multilevel models is my central course in applied statistical modeling, especially for the social sciences. If you're serious about using regression as an applied research tool, I recommend this course.
• Applied Bayesian statistical computing is a course in applied statistics from the perspective of programming, rather than models and algebra. I recommend it to statisticians and also applied researchers who write programs to analyze data. The course might also be of interest to computer scientists.
• Bayesian data analysis covers Bayesian data analysis from first principles. The course is required for first-year Ph.D. students in statistics, and I recommend it to anyone who wants to learn Bayesian statistics and can handle the mathematics (integrals, derivatives, and computer simulation, but no theorems and proofs).
• Research in quantitative political science is for students who are working on research projects, or who want to work on research projects, with our group on quantitative social science. You can show up to meetings if you're not sure whether you want to join in.
• The teaching of statistics at the university level provides instruction, practice, and feedback and teaching. The course is required for Ph.D. students in statistics, and I recommend it to anyone who is planning to teach statistics or related quantitative methods.
• Decision analysis is not part of any "sequence." Anyone can take the course. I recommend it to political scientists, psychologists, economists, and engineers who are interested in the theory and practice of optimal decision-making. The course is also a great way to learn about an important area of application of probability theory.
• Sample surveys is not part of any "sequence." Anyone can take the course. I recommend it to political scientists and others who are planning to conduct surveys or analyze survey data, and to statisticians who want to know about these important tools.

Back to Andrew Gelman's homepage