dissUtils (version 1.0)

diss: Many Different Ways to Quantify Dissimilarities Among Multivariate Data

Description

this function will create a distance object corresponding to the dissimilarities between rows in a matrix X, or a matrix of dissimilarities between the rows of matrices X and Y

Usage

diss(X, Y = NULL, method = "euclidean", init.info = NULL)

Arguments

X

a matrix of numeric data

Y

a second matrix of numeric data, which must have the same number of columns as X

method

a character string that uniquely matches one of the following:

braycurtis

Bray-Curtis difference, should use proportions

canberra

Canberra difference, should use proportions

chebyshev

Largest difference in any one dimension, like in chess

covariance

You may want to transpose the data before using this

euclidean

multivariate 2-norm

equality

the sum of exactly equal elements in each row

hellinger

Hellinger difference

jaccard

Jaccard distance

mahalanobis

Euclidean distance after scaling and removing covariance, which you can supply with init.info

manhattan

The sum of each dimension, no diagonal movement allowed

minkowski

arbitrary n-norm, so that init.info=2 yields "euclidean" and init.info = Inf yields "chebyshev" (but don't do the latter!)

pearson

Pearson product-moment correlation, you may want to transpose the data

procrustes

Doesn't scale or rotate, just treats the vectors as matrices with init.info columns and calculates total distance between homologous points

init.info

some methods require additional information. see above

Value

if is.null(Y), returns a distance object containing pairwise dissimilarities between the points in X.if is.matrix(Y), returns a nrow(X) by nrow(Y) matrix containing pairwise dissimilarities between each point in X and each point in Y.