PSYCH 100A Chapter 14: Psych 100A Ch14
Document Summary
Chapter 14 inferences about two or more means: analysis of variance. Analysis of variance (anova): testing hypotheses about two or more means. 14. 1 why multiple t tests are not appropriate. Pairwise null hypothesis: compares one mean one another. Multiple t tests have inflated experimentwise probability of a type i error. Increased probability of incorrectly rejecting the null when it is true. Experimentwise = 1 - ( 1- )c where c = number of null hypotheses tested. The tests are not in general independent so experimentwise is usually smaller than what we calculated. H0: 1 = 2 = 3 = 4. Complex null hypothesis: compares some combination of means with another combination of means. Omnibus hypothesis: implies that all possible pairwise and complex null hypotheses. At least one of the possible null hypotheses is false. We reject the omnibus null if any one of the pairwise or complex expressions is false. Anova does not tell us which null hypothesis is false.