I. Read a book
Read How to Lie with Statistics, by Darrell Huff. You can easily buy a copy (It’s like $13), but the library has several copies that you can check out from reserve. The original was published in 1954, and the reprint in 1993, but they are identical except for the cover. Be sure to enjoy the 50’s illustrations, and the old-fashioned portrayal of women (apparently, all women have vacuum cleaners). Much of it was covered in chapter 1.5
II. Develop a Lie
After you have read this book, you and your one or two partners (groups should have 2-3 members) should find some real world data that do not look very interesting. That is, there should be no relationship at all, the relationship is weak, or the relationship is boring. The library has several books of data, and the internet is just full of such things. I’ll give you a list of excellent sites in class. If you are stuck, come see me. You should tell me your topic by early November.
III. Write a paper based on the lie
Using the techniques learned in the Huff book, write a three-to-six page paper (plus graphs and statistics things) that demonstrates something interesting (and not true) about the phenomena, including bad assumptions, incorrect statistical analysis, misleading graphs, pie charts, etc. Creativity will be rewarded, although you should be sure to know the difference between the kinds of lies described in the book, and avoid actual lies, committing math errors, etc. (e.g. Using the mean when the median is more appropriate would be excellent. Making up a mean, because the one from your data isn’t as good, is just wrong.).
IV. Come Clean
Write another three-to-six page (plus graphs and statistics and things) Appendix where you announce your dishonesty to the world. Grade-wise, this section will be at least as important as the actual body of the paper. In the appendix, do the correct (but probably less-interesting) description and analysis of the data, including correct statistical analysis and graphs (this one had better not be a pie chart). Use at least one appropriate statistical test. If you don’t know which one to use, ask me. Point out what you did in the first paper to draw your conclusions and why that was wrong.
My preferred format:
- Cite everything, including the Huff book (in a correct citation form).
- Include a single title page, with your names, dates and course information on it. Do not put your names on any of the other sheets of paper.
- Be sure that your papers are well written with good grammar, and logical flow.
- The first part of the paper should be written in a fun, interesting format. You might make it sound like (maybe even look like) a newspaper or journal article. You might write in the style of a formal journal article with a methods section, etc., but a newspaper article or other format could also work well. Get Creative. Learn how to use all those features in Microsoft Word.
- The second part of the paper should be written in a much more serious style. Either write in in a strict research report format (where you respond to the claims in the first part), or some other style of response, like an editorial to a newspaper, a co-worker examining the first work, or a faculty member grading the paper.
- I have samples for you to look at in my office
- I imagine that each group will come talk to me at least once before you get started on the paper.
Topic proposals (a list of group members and a sentence or two, via e-mail) are due by Tuesday, November 13.
Give your proposal e-mail the subject line “Stat 190 Project Proposal”
The final paper is due at the beginning of class on Monday of the last week of class, December 3.