R help - Aug 2013 - caTools AUC different between mutli-class and subset binary

Greetings,

This is more of an explanation question but I was using the colAUC function
on the iris dataset and everything works smoothly.  This provides the AUC
for each pairwise comparison.  I decided to do the actual subset for one of
the comparisons and the numbers are different (.9326 v. .9152).  How/Why
could this be?

require(caTools)
data(iris)

colAUC(iris[,1], iris[,5], plotROC=FALSE, alg = "ROC")
> colAUC(iris[,1], iris[,5], plotROC=FALSE, alg = "ROC")           
[,1]
setosa vs. versicolor    0.9326
setosa vs. virginica     0.9846
versicolor vs. virginica 0.7896

set_vers <- subset(iris,
Species==c("setosa","versicolor"))
set_vers$Species <- factor(set_vers$Species)

colAUC(set_vers[,1], set_vers[,5], plotROC=FALSE, alg = "ROC")
> colAUC(set_vers[,1], set_vers[,5], plotROC=FALSE, alg = "ROC")   
[,1]
setosa vs. versicolor 0.9152


Regards,

-- 
Charles Determan
Integrated Biosciences PhD Candidate
University of Minnesota

	[[alternative HTML version deleted]]