ENVCLASS

Introduction

Contact : Ton Snelder

EnvClass is a set of functions based on the R statistical language and environment. It helps in both the development and analysis of environmental data classifications. The analysis functions are based on the description of combinations of objects (site) by multivariate data and they include homogeneity analysis, mean similarity analysis, the Mantel test and analysis of similarities (ANOSIM).

EnvClass provides options that supplement the use of these tests such as simultaneous calculation of the necessary dissimilarity and statistical test matrices, automatic extraction of classes that do not have a sufficiently large predefined sample size, and calculation of confidence intervals and p values. EnvClass also contains utility functions for testing several classifications in the same operation and which also provide supplemental information on the performance of the classification by comparing the statistical tests obtained by classification with the statistics resulting from the clustering of the data itself (a.k.a. a posteriori classification) and/or obtained by grouping the sites based on the criterion of their geographic location. 
 
EnvClass contains five functions for defining the classifications from the "bottom up" (i.e., by combining the independent variables a certain way).
These methods are based on correlation of the dissimilarity matrix, modelling of dissimilarities, the classification and regression tree (CART), linear discriminant functions and Random Forests.