Prediction distances


I performed a PLS-DA and then I would like to predict new observations. Indeed, I could predict the classes, but there are some aspects that I don’t how interpret them.

Normally in literature, prediction probabilities are mentioned as a measure of how well an observation is classified or simply to know the probability of an observation to belong to a specific class.

Predict results from Mixomics don’t give probabilities, but prediction distances. How could I interpret such distances? I know the function offers three distances, but my question is about what do they mean for each observation?

Thanks very much!!


Hi @enzo,

See section 1.3 in S1 of the mixOmics article: