STA 106 Study Guide - Final Guide: Type I And Type Ii Errors, John Tukey, Analysis Of Variance
Document Summary
With anova, we usually want to make more than ci. 1 2, 1 3, 2 3. Typically, you decide what cis you are interested in au priori. Issue when we make k cis is that overall type i error increases. = prob of random sample from the tail. 1, 1 2, 1 overall type i prob we draw from the tails in at least one of the k cis. Suppose 90% ci for i j are individual error ( 1 2): (2,8) ( 1 3):(10,12) 2 ,nt a is replaced by (1 )100% tukey corrected ci. 2 q1 ,a,nt a where q is a studentized range distribution with df (1) = a and df (2) = nt a and t > t1 at the same df. Interpreting a corrected interval, we are overall/simultaneously/family-wise (1 )100% con dent Can be used with any contrast (pairwise too) a (1 )100% sche e interval for a ci i is.
