Hello, I using proc discrim for classification purposes.
When I use method= normal, proc discrim give the classification function (also know as Fisher´s linear discriminant function), and then I can classify new observations into the groups in a paper and pencil way (not need used software to do this).
But when I used method = NPAR (non-parametric) SAS don´t give the fisher function, and it used Kernel or K-Nearest Neighbor, and the I to classify new observations I need to rerun the proc discrim again with the testdata option.
My doubt is: is there other way (manually) to classify news observations using non-parametric method? If so, what is?
I am using NPAR because it result better predictive results than normal.