Hi all,
I have a multicategory classification system and I am looking for
some reliable accuracy stats. I'm used to binary systems and usually
use total percent correct, sensitivity, and specificity, and along
with the ROC AUC stat. Are you aware of good accuracy stats for
multicategory systems?
As an added complication, the training sample is unbalanced, so I can
achieve 77% correct (the max percentage correct) by always choosing
option #1. Therefore, I need something that takes this into account
and is chance-corrected. Any ideas?
Best,
Dan
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|