🚀 go-pugleaf

I work in a clinical research setting where we frequently examine
parameters obtained during exercise stress testing in an effort
to develop methods that will permit an increased sensitivity and
specificity for the prediction of coronary artery disease.  Our 
usual approach is to apply two different variables or sets of 
variables to the same sample, and then ask which provides the 
better diagnostic utility.

I have been structuring an agreement matrix with the cells
containing those patients correctly and incorrectly classified by
each variable set; I then test for a difference in the discordance 
using McNemar's test.  I am also looking at comparing the areas 
under the two ROC curves.

I would appreciate any suggestions of any other methods I should
be considering.

Thank you,

Randy Crump