Description of Term Project for STAT 529

The project can be more-or-less anything related to the course. If you have dissertation or RA projects that involve sampling you are positively encouraged to use them.

All projects must be passed by Professor Handcock before they are acceptable.

Possibilities include

• A review of some area we didn't cover enough in class.
• Design of a survey
• Data analysis of a complex survey sample (and the impact of the design on the analysis)
• A simulation study or mathematical investigation of some question.
• Teaching materials for explaining some important or subtle issue
• Comparison of graphical approaches to presenting or exploring data

I do not want a quick and routine analysis; a good job will show understanding of the problem and possible solutions and techniques to consider. The technical results should be stated clearly. The report must contain a clearly written conclusion section giving a non-technical summary that is concise and informative.

The projects will be presented in poster sessions during class time (10:30am-12:00pm) on June 4, Gowan 1a. I will invite Stat, Biostat and CSSS faculty and students.

You can also hand in supplemental material in the form of a manuscript to support the analysis on the poster. This will be due Tuesday 10:30am, June 11th

There is no need to get a fancy printed poster (unless you want to practice fancy printed poster design, a useful skill), but it is important to design the poster so that it can be read and understood by your target audience.

The project will be 40% of the grade for the class, with 30% homework and 30% midterm exam.

Specific suggestions:

Reviews

• Some topic in network sampling
• Line/transect/quadrat sampling
• Internet sampling: threat or menace?
• Small-area estimation
• What survey techniques are used (and whether they are used appropriately) in some field of application.
• What do Bayesians do about complex samples?

Simulation or analytic studies

• Sequential stratification (as suggested in class)
• Efficient designs for two-phase sampling
• Impact of omitting weights from regression models
• Two-phase sampling for handling imputed genes in whole-genome association studies

For data analyses or designs, if you talk to me about what area you are interested in, we can try to find suitable data.