Statistical Sciences 2244A/B Lecture Notes - Lecture 18: Type I And Type Ii Errors, Multiple Comparisons Problem, Null Hypothesis
15 Jan 2017
School
Department
Professor
Document Summary
Lecture 018 - multiple comparisons (post-hoc tests) - logic and. Procedures that control the overall (i. e. family) type i error rate in multiple comparison situations. Once in awhile, we get a sample that rejects the null. The null is that all the sample means are the same. Rejection of th null based on a single population test is more likely to closely reflect the real state of the. Rather than doing 20 tests can getting a rejection simply because it was an odd sample (due to repeated samples) The significance level is adjusted for each comparison based on the overall significance and the number of pair wise comparison being made. Increasing the p-value to make it harder to reject the null hypothesis. In a 2-sample pooled" 1-test (after significant anova) Estimated standard deviation of differences in means. Assumes constant variance ( i ) for the test statistic (df(error)) Anova assumes a constant variance among the various populations under study.
