BIOL 3P96 Lecture Notes - Lecture 15: Data Dredging, Alternative Hypothesis, Random Effects Model
25 Jul 2015
School
Department
Course
Professor
Document Summary
Biol 3p96 lecture 15: comparing means of more than two groups. Carrying out many statistical tests in hope of finding at least one significant result. The probability of making at least one type i error is greater than the significance level when many tests are conducted. Probability of making no type i errors in n independent tests is (1 - )n. 0. 95 with one test, 0. 77 with five tests, 0. 60 with 10 tests, 0. 36 with 20 tests. Probability of making at least one type i error is 1 (1 - )n. Simplest way to be rigorous when conducting multiple tests, is to correct for the inflation of type i error rates. Uses a significance level of /number of tests. E. g. for 12 tests, use an of 0. 05/12 = 0. 00417. Probability of getting at least one type i error remains equal to 0. 05. However, the power of the test is reduced. Carries out all tests at = 0. 05.
