Hello,
I'm looking for a way to test a different LDA strategy than the standard One-off (or hand-out) method when trying to predict a categorical response with multiple classes.
I'm particularly interested in the One-Versus-Rest (OVR) (also referred to as One-Versus-All or OVA) and the One-Versus-One (OVO) strategies.
So, instead of using the default One-off LDA which seeks to solve the classication problem for all classes in the model simultaneously I would like to try testing OVR strategy (which divides each class into two groups, trans- forming the k-class classi?cation problem into a k − 1 round binary classi?cation problem) and OVO strategy.
Any idea how to do this in JMP?
Thanks in advance!
