Taxonomic Application of Classification and Regression Tree (CART) and Random forests (RF): a Case Study of Middle Cambrian Trilobites

Abstract

The morphological variables often have non-normal distribution The statistical analysis of such variables by imposing normal assumption invariably yields unreliable results Classification and Regression Trees (CART) and Random Forests (RF) are non parametric techniques that are alternative to conventional classification methods such as cluster analysis and linear discriminant analysis used in morphometric research This paper uses the aforementioned Non-Parametric techniques to the variables of the cramdial features of the trilobite genera Hundwarella and Iranoleesia. It is found that misclassification rates in CART and cluster analyses are comparable, whereas they are reduced substantially by the use of Random Forests.