Statistical Sciences 2244A/B Lecture Notes - Lecture 19: Type I And Type Ii Errors, Bonferroni Correction, Multiple Comparisons Problem
24 Dec 2016
School
Department
Professor
Document Summary
Bonferroni correction: divide alpha by the number of pairwise comparisons we are performing, we are thus being stricter for each test. Or calculated adjusted p-values by multiplying p-value by # pairwise comparisons: alternative method is to increase p-value making it harder to reject null hypothesis. For defoliation and species richness example, adjusted = Note: we must do a two tailed alternative hypothesis. Clicker: anova assumes a constant variance among the various populations under study. Under the model there is a constant variance ( !) within groups ppooled and ms(error: estimated using deviation of all values around their respective treatment group mean (hence, df(error) = n k, therefore, estimated by ms(error) Now we can do a pooled t-test since we are assuming the variance is the same between the two populations anyways. The same equation as the other two sample independent test modified since we can now pool our variance.
