Stat 4893W Information Fall 2007

Stat 4893W students do not meet together on a regular schedule. Instead they are should remain in frequent contact with their adviser — the Statistics faculty member who is supervising their paper as they develop an outline, collect references and write as many drafts as are necessary to get a final paper.

The Director of Undergraduate Studies (Kit Bingham) is the instructor of record and is responsible for turning in grades. The faculty member supervising a paper informs him of the grade.

You should choose the topic of the paper in consultation with a Statistics faculty member who agrees to supervise your work. Often the faculty member will be your instructor in a 5xxx Statistics course and the material will build on material from the course. The Director of Undergraduate Studies is the default supervisor.

You should be familiar with the description of Stat 4893W on the web.

Schedule

Here is the expected time table:

If your paper is supervised by another faculty member, she/he may have a different schedule.

Drafts may be submitted as Microsoft Word® documents. Be advised, however, that such documents prepared on a Windows computer often are not properly displayed on a Macintosh which is what I use. This is particularly true of Greek letters and complicated formulae. Usually, a PDF document created from a Word document works fine on a Macintosh. For complicated formulae, the Equation Editor which is an add on to Word is wonderful.

It is important to keep in touch with your faculty supervisor while working on the paper. Unless you are told otherwise, you will be expected to provide progress reports and/or drafts at intermediate stages. Presenting a “finished” product without consultation is risky.

Format of paper

The outline and all drafts must be dated and have page numbers. Double spacing drafts is a good idea.

The final draft should have the following:

• A title page including the title of the paper, your name, the course number, the supervisor of the paper and the date.
• An abstract succinctly describing what the paper is about. This should be on the order of 10 lines and can appear on the title page, on a separate page or before the first paragraph of the paper proper. It normally has considerably wider margins than the text of the paper.
• On the order of 10 to 15 pages of text, excluding figures and tables; this is not a hard and fast requirement. This can be single or 1.5 spaced, but put an extra blank line between paragraphs. Any computer output should be in a monotype font such as Courier.
• A list of references at the end in one of the two formats described below.

Formats for references

References should be alphabetized by author or, in the case of internet references, by organization hosting the information. The list should be in one of two formats.

Numbered:

[1] Anderson, T. W. An Introduction to Multivariate Statistical Analysis, Wiley, New York (1958)

[2] Andrews, D. F., Gnanadesikan, R, and Warner, J. L. Transformations of multivariate data. Biometrika 27 (1971) 825-840

[3] Farlie, D. J. G. The performance of some correlation coefficients for a general bivariate distribution Biometrika 47 (1960) 307-323

You reference these in the text by number, possibly with a page number or equation number, as in

The conditional distributions also agree closely with normal theory [24]. but the correlation coefficient ….

Anderson [1, p. 37] considers a bivariate normal distribution ...

Unnumbered:

Abramovitz, M. and Stegun, I. A., 1964. Handbook of Mathematical Functions. Washington, DC: U.S. Government Printing Office.

Benton, D., Krishnamoorthy, K. and Mathew, T., 2002. Inferences in multivariate-univariate calibration problems. Revised version submitted to The Statistician (JRSS-D).

Chattamvelli, R. and Shanmugam, R., 1997. Computing the Noncentral Beta Distribution Function. Applied Statistics, 46, 146-156.

Craig, C. C., 1942. Note on the distribution of noncentral t with an application. Annals of Mathematical Statistics, 17, 224228.

You refer to these in the text by author's name and date, possibly with a page number or equation number, as in the following:

Other uses of the noncentral t distribution include computing confidence intervals and hypothesis tests about the independent variable in a multivariate-univariate calibration problem (Benton, Krishnamoorthy, Mathew 2002).

Further, we have the following well-known recursion relations (see Abramovitz and Stegun 1964, 26.5.16) for the incomplete beta function in (2.4):

Reference to articles on the web are acceptable as long as they include a complete URL, the date they were referenced and the author. However, as you know, there is a lot of inaccuracy on the web, particularly in Wikipedia. Whenever possible you should verify information from a book or article.

The JSTORE web site includes PDF versions of hundreds of Journals. Even more are accessible on line through the University library.

If you have trouble getting started, you may find help from a writing consultant in the Center for Writing’s Student Writing Support program (SWS). You can get help at all stages of the writing process. SWS provides help in 15 Nicholson Hall (primarily by appointment) and 9 Appleby Hall (walk-in) plus some satellite locations. You can schedule an appointment online.