STAtistics / Mr. Hansen	Name: __________________________
4/18/2011	Group number: ______

Power Analysis Worksheet

1. Review all 6 groups’ research questions for suitability. Mark any required changes with standard proofreading marks (carets, deles, etc.).

Group 1: Nick R.-S., Justin, Dominique
   In an audio-only experiment, does the gender of voices affect the likelihood that STA boys consider them to be “trustworthy”?

Group 2: Alex, Andrew, Phineas
By how much does knowing the type of material that will be on a quiz or test improve the mean score?

Group 3: Daniel, Chick, Preston
Is students’ ability to distinguish between Pepsi and Coca-Cola in a blind taste test better than chance alone would predict?

Group 4: Jamie, Julien, Brennan
   Are all forms in the STA Upper School equally competent at finding Waldo? In other words, is the proportion of successful Waldo-finding within some number of seconds the same across forms?

Group 5: Edward, Zeke, Jordan
Is the mean time freshmen estimate for a blind 150-second wait larger than the mean time for seniors?
Revised version: Is the mean |error| for freshmen when estimating a blind 150-second wait different from the mean |error| for seniors?

Group 6: Nick S., Tip, Andrei, Ousmane
Is fitness (as measured by pushups, situps, hand touches, or some merged statistic) better for lacrosse/crew than for golf/baseball?
Revised version: Are “educated guesses” better than single-letter guessing on multiple-choice quizzes (matched pairs) where subjects have no actual knowledge of the subject matter?

2. What statistical test (or C.I. construction technique) is appropriate for each group?
1. ___________________
2. ___________________
3. ___________________
4. ___________________
5. ___________________
6. ___________________

3. Work with your group members to construct null and alternative hypotheses. Check with Mr. Hansen before proceeding.

_____________________________________

_____________________________________

4. Work with your group members to construct a believable table of fake data that illustrates the null hypothesis. Try to engineer your data so that the standard error of the statistic (or, for those of you who are doing chi-square tests, the relative size of counts) is believable. Show table on reverse, using the sample size that you proposed.

5. Now write your ES that you think is believable. (Note: Some of the groups forgot to do this earlier.) Answer: _________

6. Now rewrite your table from #4 so that it illustrates the ES you claimed in #5.

7. Run a full statistical test (PHASTPC, including S) to see whether statistical significance is achieved. Make a sketch to illustrate power, and shade the power. What is your power estimate? _________ Is this acceptable? ______ If not, what changes to your methodology would you propose? (Write a sentence or two.)