Identification of taxon through classification with partial reject options

Abstract Identification of taxa can significantly be assisted by statistical classification based on trait measurements either individually or phylogenetic (clustering) methods. In this article, we present a general Bayesian approach for classifying species mixture continuous and ordinal traits, any type covariates. The vector is derived from latent variable with multivariate Gaussian distribution. Decision rules supervised learning are presented that estimate model parameters through blocked Gibbs sampling. These decision regions allow uncertainty (partial rejection), so not necessarily one specific category (taxon) output when new subjects classified, but rather set categories including the most probable taxa. This discriminant analysis employs reward functions set-valued input argument, an optimal Bayes classifier defined. We also way safeguarding against outlying observations, using analogue p-value within our setting. refer to as Karlsson–Hössjer method, it illustrated original ornithological data birds. incorporate selection cross-validation, exemplified another