Chapter 12--Multisample Inference

05215 fisher confintmccalphaqt1 005615 bonferroni

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: VA) Multiple Comparisons The Kruskal-Wallis Test Two-Way ANOVA Example: R Code library(multcomp) #need to download and install the package BP<-c(27, 26, 21, 26, 19, 13, 15, 16,15, 10, 10, 11,22, 15, 21, 18,20, 18, 17, 16) T<-rep(c("S","AL","AH","BL","BH"),each=4) Tx<-factor(T) A<-aov(BP~Tx) summary(A) contr<-rbind("AL-AH"=c(0,-1,1,0,0), "BL-BH"=c(0,0,0,-1,1), "A-B"=c(0,1/2,1/2,-1/2,-1/2)) mc<-glht(A,linfct=mcp(f=contr)) summary(mc,test=adjusted(type=c("none"))) #Fisher summary(mc,test=adjusted(type=c("bonferroni"))) confint(mc,calpha=qt(1-0.05/2,15)) #Fisher confint(mc,calpha=qt(1-0.05/6,15)) #Bonferroni Chapter 12: Multisample Inference Stat 491: Biostatistics Analysis of Variance (ANOVA) Multiple Comparisons The Kruskal-Wallis Test Two-Way ANOVA Example Cont’d... Surprised by the mean drop in the blood pressure for the two treatments involving drug A, the investigator decided to see whether the data provides evidence (at α = 0.05) that the BP reductions for S+A and S are different. ˆ4 = (1/2)(¯AL + yAH ) − yS = −11.375 y ¯ ¯ and ˆ V ( ˆ4 ) = ( 1 )2 ( 1 )2 (−1)2 2 +2 + 4 4 4 Chapter 12: Multisample Inference × 6.433 = 2.412. Stat 491: Biostatistics Analysis of Variance (ANOVA) Multiple Comparisons The Kruskal-Wallis Test Two-Way ANOVA Example Cont’d... Scheffe’s Method S= ˆ (k − 1)F1−α,k −1,n−k V ( ˆ4 ) = √ 4 × 3.06 × 2.412 = 5.430. Since | ˆ4 | > 5.430, we conclude that ˆ4 is significant. Assuming the decision to test the significance of ˆ4 was made prior to looking at the data, Fisher’s method rejection cut-off point is t1−α/2,n−k ˆ V ( ˆ4 ) = 2.131 × 1.553 = 3.311. Although, both method declare ˆ4 significant, Scheffe’s cut-off point is bigger than that of Fisher’s. Hence, Fisher’s method is more likely to declare significance. Therefore, it should not be used for post-hoc comparisons. Chapter 12: Multisample Inference Stat 491: Biostatistics Analysis of Variance (ANOVA) Multiple Comparisons The Kruskal-Wallis Test Two-Way ANOVA All Pairwise Comparisons Used in experiments that are exploratory in nature. Give k treatments, there are k (k − 1)/2 possible pairwise comparisons. So if k = 6, there are 15 possible pairwise comparisons. Needless to say, Fisher’s, Bonferroni and Scheffe Method discussed previously can be used. Fisher’s will likely declare many false significance. Bonferroni and Scheffe will tend to be too conservative. Chapter 12: Multisample Inference Stat 491: Biostatistics Analysis of Variance (ANOVA) Multiple Comparisons The Kruskal-Wallis Test Two-Way ANOVA Strategy for All Pairwise Comparison (k=4) 1 Arrange the sample means from the smallest to the largest y(1) , ¯ 2 y(2) , ¯ y(3) , ¯ y(4) ¯ with the corresponding sample sizes n(1) , n(2) , n(3) and n(4) . Determine the order in which the difference will be tested, (4, 1), (4, 2), (4, 3), (3, 1), (3, 2), (2, 1) 3 The difference y(i ) − y(j ) is declared significant if ¯ ¯ y(i ) − y(j ) > LSDij ¯ ¯ 4 Group toget...
View Full Document

This note was uploaded on 02/03/2014 for the course STAT 491 taught by Professor Solomonharrar during the Fall '12 term at Montana.

Ask a homework question - tutors are online