NTHU STAT 6681
Design and Analysis of Multistratum Randomized Experiments
Feb ~ Jun 2007
Instructor
Institute of Statistics,
National Tsing Hua University
(鄭少為, 清華大學/統計所)
The course will be given on the basis of 3 hours a week (Tuesday, 1:00~4:00 PM) in 綜合三館 room 840. The materials for teaching will be regularly posted on the course website (http://www.stat.nthu.edu.tw/~swcheng/Teaching/stat6681/).
This is an advance course in experimental design. Regular factorial designs are a class of designed experiments that are commonly used in many fields. In many experiments, a complete randomization of the individual experimental units is difficult to achieve. If a factorial design is performed with restrictions on randomization, this will influence the error structure because of the appearance of some random effects. This issue must be considered in the design, analysis, and statistical modeling stages.The course will introduce a very general methodology based on the concept of stratum. The methodology can be applied on various restricted schemes of randomization, such as nested designs, row-column design, Latin square, blocked design, split-plot designs.
There is no regular textbook for the class, but extensive lecture notes will be placed on the course website as the course proceeds and prepared materials will be distributed as needed. A substantial part of the lectures will be based on a lecture note developed by Prof. Ching-Shui Cheng (Department of Statistics, University of California, Berkeley).
Bailey, R.A. (expected publication date: 2007/4), Design of Comparative Experiments, Cambridge University Press.
Milliken, G.A. and Johnson, D.E. (1992), Analysis of Messy Data, Vol. I: Designed Experiments, Chapman & Hall.
My office is in 綜合三館 room 818 (phone extension: 33162). My office hour is scheduled on Tuesday, 4:00~5:00PM. Questions by e-mail are also welcome: my e-mail address swcheng@stat.nthu.edu.tw.
Your grade will be determined by two paper presentations (60%) and one project presentation (40%) on the topics that will be posted in the future. Each oral presentation will be given by one student, unless the topic is very difficult, in which case two students can split the work. Each presentation should last 30-40 minutes. It should be given using slides created in the electronic format of your choice so that we can archive the presentations for future reference.
Knowledge of basic course in experimental design and regression analysis.