Parameter tuning and performance assessment functions in mixOmics can use three different distance metrics to predict the class of test data in supervised, classification models (e.g. (s)PLS-DA, DIABLO, and MINT (s)PLS-DA). These distance metrics are: max.dist, centroids.dist, and mahalanobis.dist. This page describes the difference between these distance metrics and demonstrates their use in examples.
Related case studies:
sPLS-DA SRBCT Case Study
References:
1. Mahalanobis, P.C. (1936) “On the Generalised Distance in Statistics.”. Sankhya A 80, 1–7 (2018). https://doi.org/10.1007