Adel
2013-Nov-29 18:08 UTC
[R] the wilcox.test() and pairwise.wilcox.test are producing different results
Dear list member, I want to compare if the rank order is significantly different for seven different measures. So we have same sample but different measures which reduces the problem to a paired one sample Wilcox test if I understood the test correctly. In constructed toy examples for my sake of understanding, but things are not adding up. Basically, the wilcox.test() and pairwise.wilcox.test are producing different results when they should not (according to my understanding of course): #take a vector daily.intake <- c(5260,5470,5640,6180,6390,6515, + 6805,7515,7515,8230,8770) #I get desired results when I do the following daily<-data.frame(pre=daily.intake, post=daily.intake) #add som differences daily[1,1]<-5000 daily[2,1]<-5100 #reshape the data for pairwise comparison library(reshape2) daily_long<-melt(daily, id=) #conduct simple test wilcox.test(daily$pre,daily$post, paired=T) #produces desired results #do the test again but now in a pairwise, which produces the same p-value as in the simple test above pairwise.wilcox.test(daily_long$value,daily_long$variable, paired=T) #But now I the issues arise when testing more than two vectors. #take three vectors this time daily<-data.frame(pre=daily.intake, post=daily.intake, posttwo=daily.intake) #add some differences daily[1,1]<-5000 daily[2,1]<-5100 daily[10,3]<-9000 daily[11,3]<-9100 #the wilcox.test() produces a set of p-values wilcox.test(daily$pre,daily$posttwo, paired=T) wilcox.test(daily$pre,daily$post, paired=T) wilcox.test(daily$post,daily$posttwo, paired=T) #and the pairwise.wilcox.test produces another set pairwise.wilcox.test(daily_long$value,daily_long$variable, paired=T) ##And from the manual for pairwise.wilcox we get similar issues #produces a given set of p-values attach(airquality) Month <- factor(Month, labels = month.abb[5:9]) ## These give warnings because of ties : pairwise.wilcox.test(Ozone, Month) pairwise.wilcox.test(Ozone, Month, p.adj = "bonf") detach() #but if we want to test the rank difference between the 6th and 7th month we get a p-value of 0.5775 testar skillnaden mellan 6e och 7e månaden – observ however that this data is not paired which makes it different to the example I gave above. #p-värdet är 0.5775 air<-subset(airquality, airquality$Month < 7) #p-value is now 0.1925 wilcox.test(air$Ozone~air$Month) What am I doing wrong here? Best Adel -- Adel Daoud, PhD Visiting researcher (post-doc) Max Planck Institute for the Study of Societies / Max-Planck-Institut für Gesellschaftsforschung Paulstr. 3 | 506 76 Köln | Germany Tel.: + 49 (0) 221 2767-534 daoud@mpifg.de Department of Sociology and Work Science, University of Gothenburg Box 720 405 30 Göteborg, Sweden Visiting address: Sprängkullsgatan 25, room K109 +46 031-786 41 73 Adel.daoud@sociology.gu.se -- View this message in context: http://r.789695.n4.nabble.com/the-wilcox-test-and-pairwise-wilcox-test-are-producing-different-results-tp4681375.html Sent from the R help mailing list archive at Nabble.com. [[alternative HTML version deleted]]
arun
2013-Nov-29 19:29 UTC
[R] the wilcox.test() and pairwise.wilcox.test are producing different results
Hi, Have you tried p.adjust="none" ?pairwise.wilcox.test(daily_long$value,daily_long$variable, paired=T,p.adj="none") pairwise.wilcox.test(Ozone, Month, p.adj = "none") A.K. Dear list member, ? I want to compare if the rank order is significantly different for seven different measures. So we have same sample but different measures which reduces the problem to a paired one sample Wilcox test if I understood the test correctly. In constructed toy examples for my sake of understanding, but things are not adding up. Basically, the wilcox.test() and pairwise.wilcox.test are producing different results when they should not (according to my understanding of course): ? #take a vector daily.intake <- c(5260,5470,5640,6180,6390,6515, ????????????????? + 6805,7515,7515,8230,8770) ? #I get desired results when I do the following daily<-data.frame(pre=daily.intake, post=daily.intake) ? #add som differences daily[1,1]<-5000 daily[2,1]<-5100 ? ? #reshape the data for pairwise comparison library(reshape2) daily_long<-melt(daily, id=) ? #conduct simple test wilcox.test(daily$pre,daily$post, paired=T) #produces desired results ? #do the test again but now in a pairwise, which produces the same p-value as in the simple test above pairwise.wilcox.test(daily_long$value,daily_long$variable, paired=T) ? #But now I the issues arise when testing more than two vectors. ? #take three vectors this time daily<-data.frame(pre=daily.intake, post=daily.intake, posttwo=daily.intake) ? #add some differences daily[1,1]<-5000 daily[2,1]<-5100 daily[10,3]<-9000 daily[11,3]<-9100 ? #the wilcox.test() produces a set of p-values wilcox.test(daily$pre,daily$posttwo, paired=T) ?wilcox.test(daily$pre,daily$post, paired=T) wilcox.test(daily$post,daily$posttwo, paired=T) ? #and the pairwise.wilcox.test produces another set pairwise.wilcox.test(daily_long$value,daily_long$variable, paired=T) ? ? ? ##And from the manual for pairwise.wilcox we get similar issues ? #produces a given set of p-values attach(airquality) Month <- factor(Month, labels month.abb[5:9]) ## These give warnings because of ties : pairwise.wilcox.test(Ozone, Month) pairwise.wilcox.test(Ozone, Month, p.adj "bonf") detach() ? #but if we want to test the rank difference between the 6th and 7th month we get a p-value of 0.5775 ? testar skillnaden mellan 6e och 7e m?naden ? observ however that this data is not paired which makes it different to the example I gave above. #p-v?rdet ?r?0.5775 ? air<-subset(airquality, airquality$Month < 7) ? #p-value is now 0.1925 wilcox.test(air$Ozone~air$Month) ? ? What am I doing wrong here? Best Adel -- Adel Daoud, PhD Visiting researcher (post-doc) Max Planck Institute for the Study of Societies / Max-Planck-Institut f?r Gesellschaftsforschung Paulstr. 3 | 506 76 K?ln | Germany Tel.: + 49 (0) 221 2767-534 [hidden email] ?Department of Sociology and Work Science, ?University of Gothenburg Box 720 405 30 G?teborg, Sweden Visiting address: Spr?ngkullsgatan 25, room K109 +46 031-786 41 73 [hidden email]