WinISI discriminant Equation

kongfanli's picture

Hello  everyone 
I want to develop a discriminant Equation by WinISI, but I don’t understand “uncertainty factor” and what is the number should be chose, it is said the best is 2.5 from WinISI helpfile, when I choose the smaller number, the higher accuracy, can you tell me why? Thank you very much for your help!   

jmhhierro's picture

Hello kongfanli,
For any sample the discriminant value can be predicted according to the PLS-DA model. There are a number of criteria for deciding which class a sample belongs to but the simplest for a two class model is to set up a cut-off threshold at c=0, if positive it belongs to the “in group” of class A, if negative to the other class, and this determines the boundaries between the two classes. If you add an uncertainty factor value you can correct for an unknown samples. The bigger the uncertainty value, the bigger the area for unknown samples.