[R] p-val issue for ranked two-group test

Laurel Klein Serieys laurelklein at ucla.edu
Thu Oct 20 19:17:40 CEST 2011

I'm wondering if anyone can help me with my code.  I'm coming up dry  
when I try to get a p-value from the following code.  If I make a  
histogram of my resampled distribution, I find the difference between  
by groups to be significant.  I've ranked the data since I have  
outliers in one of my groups.

mange= c(35,  60,  81, 158, 89, 130,  90,  38, 119, 137,  52,  30,   
27, 115, 123,  31, 124,  91)

healthy= c(46, 50, 30, 58, 32, 42, 42, 33, 19, 42, 30, 26, 38, 23, 16,  
28, 42, 42, 33, 35, 51, 31, 39, 40 , 42, 38, 36, 39, 38)


exptdiff <- mean.mange - mean.healthy #the expected difference between  
between the mean of the ranked groups

both.chemistry<-c(mange, healthy) #concatenate two vectors into one in  
preparation for resampling the data

both.ranks<-rank(both.chemistry) #rank combined data in the case that  
there are outlying values in the data or the dataset is small


z<-rep(NA,reps) # z will the the simulated storage value for the  
resampling efforts

for(i in 1:reps){ #create the loop

x<- sample(both.ranks, length(both.ranks),replace=FALSE) #instructions  
for how to resample where sample the entire combined data without  

p.mange<-mean(x[(1:l.mange)])  #create a simulate mean value for the  
resampled mange values
p.healthy<-mean(x[(l.mange+1):(l.mange+l.healthy)])  #create a  
simulated mean value for the resampled healthy values

pdiff<- p.mange-p.healthy #the simulated difference between groups

z[i]<- pdiff  #the stored list of simulated differences
p=mean(z>=exptdiff)*2 #2-tailed test multiply by two

hist(z, xlab="Resample Values", main="Distribution for Two-Group BUN Test")
confints=quantile(z, c(0.025,0.975))
abline(v=confints, col="blue") #draw a line for each cutoffs
abline(v=exptdiff, col="red")


More information about the R-help mailing list