OUTLINE FOR CHAPTER
8
Effect Size, Power, CIs, and
Bonferroni
 The SevenStep Version of Hypothesis Testing
 Introduction:
 A brief review of the simplest version of hypothesis testing
 The 7th step: After rejecting Ho,
determining the degree to which Ho
was wrong
 The difference between "statistical significance" and "practical
significance"
 Two ways researchers can do Step #7
 They can compute a measure of "effect size"
 They can conduct a post hoc "power analysis"
 The NineStep Version of Hypothesis Testing
 A simple listing of the nine steps . . . with the 3 "new" steps
located in positions 4, 5, and 6
 Step #4: Specification of the effect size (ES):
 ES: the point that separates cases where Ho
is false by a small and trivial amount vs. cases where it's
false by a big and noteworthy amount
 Two options for ES: "Raw" or "standardized" (and Cohen's "standards")
 Step #5: Specification of the desired level of "power":
 The notions of "statistical power" and a "beta error"
 Why researchers donšt set power at .999
 Step #6: Determination of the needed sample size:
 Computing n from a formula or looking up the needed
n in a chart
 What to do if there is a fixed n
 The primary advantage of using the 9step version of hypothesis
testing
 Hypothesis Testing Using Confidence Intervals
 Using a confidence interval to test a null hypothesis about a
single population
 Using a confidence interval to test a null hypothesis involving
two populations
 Adjusting for an Inflated Type I Error Rate
 The notion of an "inflated Type I error rate"
 Why the Type I error rate becomes "inflated" if a fixed alpha
is used with multiple tests
 The Bonferroni adjustment technique
 The experimentwise error rate
 The DunnSidak modification
 A Few Cautions
 Two meaning of the term "effect size"
 "Small," "medium," and "large" effect sizes
 The simplistic nature of the 6step version of hypothesis testing
 Inflated Type I error rates
