Statistical Consulting Projects

There are three major data analysis problems for which written reports are required in Stat 998. For the first two, the client is interviewed in class, and each student analyzes the data and writes a report. A final (major) project consists of an separate consulting experience for each student with a required oral presentation and written report. Here are some guidelines on the scope of these projects.

Guidelines for Major Projects

Projects involve collaboration with "live" scientists. The problem itself should be primarily statistical, not computational. For example, the scientist is responsible for preparing the data in an electronic form, and the statistician focuses on issues of design, graphical presentation of data, data analysis and interpretation of results. Scientist and statistician typically work together to create a 2-page description of the problem, which in some sense serves as a "contract" defining the scientific questions and the scope of the statistical investigation.

Scientists for the first two projects come to class twice, usually a week apart, for the first two projects. Students meet individually with the third scientist, whom they choose with guidance from the instructor. All projects are completed with a 12-page report. Students give an oral presentation for the class on the third project. The goal here is to obtain feedback from the class, rather than to present a finished product. Your presentation should include a few key graphs and some remarks on design and analytical strategy.

Remember: This is a Class Project

There should be a clear understanding between scientist and statistician of the nature of the involvement. This is an assignment for a course with a definite time limit. For example, if you do an analysis for the scientist, and that person does not like the analysis, then it is not your responsibility to modify or repeat it. Moreover, if you want to continue working on the project after the course, that is your choice (you might consider asking for money and/or co-authorship). However, you do not have to continue.

Problem Definition and Shaping Analysis

One of the difficult aspects of this assignments is problem definition and shaping analysis. Feel free to ask for my help from the instructor if you are unsure how to proceed here. The process of preparing a 2-page summary handout can help you clarify these issues.

The problem should be "not too big, not too small". Simple regression or one-way anova is too small. Thousands of observations on hundreds of variables may be too big. Ideally the problem has a few non-standard twists involving missing data or imbalance. Or the research questions may be non-standard in some creative way. In other words, the problem should be sized so that it can be completed in three "leisurely" weeks during the semester. Remember that the Statistics MS exam allow only about three days per problem!

12-Page Written Report

Students must submit a 12 page written report for each project. Key graphs in the report must be closely connected to text material. See the LaTeX Sample Outline for further suggestions on report style and content.
Brian Yandell (