Experimental Design and Analysis
CIS*6650 (Winter 2010)

Instructor: Fei Song (Thornbrough 1388, ext 58067)
Email: fsong@uoguelph.ca

Office Hours: Monday and Wednesday, 4:00 – 5:00 pm or by appointments.

Lecture time and location: Monday and Wednesday, 2:30 – 4:00 pm in Reynolds 219

Course description

This course provides the students with an understanding about the design and analysis of experiments and how these skills relate to research in computer science in general and their areas of study in specific.

Pre-requisites: Basic knowledge of probabilities and statistics at an undergraduate level should be helpful, although not necessarily required.

Marking Scheme

Problem Sets 5 x 7% = 35%

Project Phase 1 (Literature Survey) 10%

Project Phase 2 (Study) 15%

– Design (5%)

– Result/Analysis (10%)

Project Phase 3 (Experiments) 30%

– Design (10%)

– Seminar (5%)

– Results/Analysis (15%)

Final exam 10%

Ideally, the project should be directly related to the student’s thesis area, but suggested projects are available for those who do not yet know their thesis topics.

Lecture Topics

· Research as “asking important questions (and answering them)”

· Literature survey, “lab” notes and experimental write-ups

· Models, Hypotheses, Theories and Experiments: the scientific method, hypotheses and experiments, models and theories, measures and probes, control of variables, and avoiding design “mistakes”

· Statistics

- Cookbook approach with an understanding of “why” in order to know which techniques to apply as well as the limits of applicability for the various techniques

- Topics: random variables, averages and variances, confidence intervals, comparisons and the T test, central limit theory, non-parametrics (rank based statistics), regression, polynomial regression, multi-factorial designs, post-hoc corrections (Bonferonni), and ANOVA (optional).