LDA: ordinal vs. categorical response
7 views (last 30 days)
Hi, I would be interested if there is a difference in applying a linear discriminant analysis if the response variable is ordinal (class 3 is better than class 2, class 2 is better than class 1) rather than categorical?
I couldn't find any good advice on that. Thanks for help
Ilya on 27 Apr 2015
There is a difference. This choice, ordinal vs nominal, should be driven by what your response variable represents.
If you have a sufficiently recent version of the Statistics Toolbox, use the fitcecoc function for ordinal labels. Set 'Coding' to 'ordinal' and 'Learners' to 'Discriminant'. For categorical (nominal) labels, simply use fitcdiscr; class labels are treated as nominal by default.