rankindex.new: Compares the efficacy of resemblance measures for a particular data scenario.

Description

This function is a wrapper for rankindex from vegan.

Usage

rankindex.new(grad, veg, indices = c("matching", "rogers", "jaccard.pa", 
"sorenson", "kulkczynski.pa", "ochiai", "gower", "steinhaus", "kulkczynski.q", 
"jaccard.q", "euclidean", "rel.euclidean", "manhattan", "czekanowski", 
"whittaker", "canberra", "chi.metric", "chi.dist", "morisita", "morisita.horn", 
"minkowski", "mountford", "raup.crick", "binomial", "chao"), stepacross = FALSE,
 method = "pearson", ...)

Arguments

grad

An s x e matrix of environmental variables, where s = the number of sites and e = the number of environmental variables.

veg

An s x p community matrix, where s = the number of sites and p = the number of species.

indices

Some subset from the character string:

c("matching", "rogers", 
  "jaccard.pa", "sorenson", "kulkczynski.pa", "ochiai", "gower", "steinhaus", 
  "kulkczynski.q", "jaccard.q", "euclidean", "rel.euclidean", "manhattan", 
  "czekanowski", "whittaker

stepacross

Logical, specifies whether a step across transformation should be applied to the resemblance matrix.

method

What method is used to calculate correlations between environmental and communtiy matrices. Must be one of "pearson", "kendall", or "spearman".

...

Other parameters to stepacross from vegan.

Value

The function returns a table of ranked Pearson's correlations (default) and a barplot.} references{Oksanen, J., Kindt, R., Legendre, P., O'Hara, B., Simpson, G. L., and Stevens, M. H. H. (2008) vegan: community ecology package. R package version 1.13-0. http://vegan.r-forge.r-project.org/

Details

A number of ecologists have evaluated the performance of distance measures by comparing environmental distance (i.e. differences along gradients) to distance in ordination space (Beals, 1984; Faith et al., 1987; De'ath, 1999). If species distributions are well described by environmental variables, then a strong association should exist between distances in species space and differences in environmental space. Thus, poor correlations represent poor performance by the distance measure. Obviously this analysis becomes more meaningful as the number of important environmental variables increases in ones measure of environmental distance. The library vegan has a function called rankindex which ranks dissimilarity or distances used for finding community distances or dissimilarities by how well these indices agree with gradient differences. The gradient separation between each point is expressed as Euclidean distance for continuous variables and as Gower's metric for mixed data (i.e. when at least some environmental variables are categorical or ordinal). In the later case the library cluster is required. The association of community and environmental distance matrices is simply the correlation of the community and environmental distance ranks and can be measured with any of the conventional measures described in Ch. 11. The function rankindex.new is a wrapper for rankindex and uses Oksanen's method to compare the efficacy of 25 of the 26 indices generated by get.dist. Mahalanobis distance is left out, since it does not create a distance matrix per se, but a simultaneous comparison of each site to all other sites.

Examples

Run this code

library(vegan)
data(varechem)
data(varespec)
r<-rankindex.new(scale(varechem),varespec)