Dear Allstat members:
Marti J. Anderson proposes a new method for non-parametric multivariate analysis of variance based on pertmutation tests. These are based on measure of dissimilarity. The author proposes 15 measures of dissimilarity: Bray-Curtis, Square root of Bray-Curtis, Euclidean, Orloci's cord, chi-square metric, chi-square distance, Hellinger, Gower, Gower (excluding double zeros), Canberra, Square root of Canberra, CY, deviance based on binomial, deviance per observation, and Kulczynski. For the ecological purposes, the author as well as Legendre & Legendre (1998) extensively use Bray-Curtis dissimilarity.
Question: What would be the appropriate measure of dissimilarity to use for general purpose (in contrast to ecological application)?
I can not find answer in Legendre, P. and Legendre, L. (1998). Numerical ecology, 2nd English edition. Elsevier. Amsterdam.
Any suggestion to help understanding the use of each of these distance measures is very welcome.
Thanks and have a very happy holiday season,
Ray Haraf
|