Homework | Due Tuesday Dec 3 2002
Download the following three scripts:
powercorrelation.SSC
powerttest.SSC
powerANOVA.SSC
[Note: To download these scripts, follow the same directions explained for other
scripts on the main class web page.]
These scripts are designed to allow you to simulate the statistical power of a test of a correlation, a t-test for the difference between two means, and an ANOVA for a 2 by 2 between-subjects design.
I would like for you to do several things for this assignment:
First, explore these scripts, try to understand how they work, and study the way in which statistical power varies as a function of sample size, the effect size, etc.
Second, for two of the three research articles that we have read and critiqued in class, conduct "retrospective" power analyses of some of the main tests. For example, given McLeod's sample size in Experiment 1 and assuming his observed means are equal to those in the population, what was the statistical power of some of his tests for Experiment 1? (I can't recall off the top of my head whether Experiment 1 used a between-subjects or within-subjects design, but let's assume the design was between-subjects for the purposes of this homework.)
Third, turn in a typed report that summarizes the power analyses you conducted (i.e., which tests from which articles and the corresponding power results) and discusses (a) whether you think the author(s) had sufficient power to detect effects (if we assume the effects were there to be detected) and (b), if power seems insufficient, how many subjects would be necessary to obtain a power of, say, 80%.