Stat Chat for 26 Jan. 2010

Agenda

• 6:00 - 6:30, Dinner


• 6:30 - 6:40, Resources on the Web: Statistics Online Computational Resource presented by Bob delMas


• 6:40 - 7:10, Journal club. George W. Cobb and David S. Moore, Mathematics, Statistics, and Teaching American Mathematical Monthly 104(9):801-823) April Kerby will moderate the discussion. Chris Malone selected the reading.


• 7:15 - 8:00, Main Event: Danny Kaplan, Hypothesis Testing: What's the Alternative?
  I know, you're thinking this is going to be a session about Bayesian inference. But it's just good ol' hypothesis testing.
  In looking through standard statistics texts and comparing what they say to what I do in my introductory classes, I've been struck by the very limited role that the alternative hypothesis plays. The alternative seems to be entirely about whether to do a one-sided or two-sided test, and is written in very abstract terms, e.g., H_a : p > 0.5.
  In my classes, I have the students work with very specific alternative hypotheses, e.g., p = 0.70. This has several advantages. It makes it clearer to the student that both the null and the alternative are *hypotheticals*, statements about world that might or might not be true. It also relates statistical practice more closely to scientific practice, where a person designing a study has to think about how big is the effect he or she is looking for. This leads directly to calculations of sample size and power. I'll show some settings in which I have students do these calculations.
  Here are some **DRAFT** slides for the presentation. Feedback/corrections/suggestions/critiques will be helpful in refining (or potentially rejecting) this work.