A Comparative Study of Permutation Tests with Euclidean and Bray-Curtis Distances for Common Agricultural Distributions in Regression

Abstract

This study describes the efficiency of permutation test for Normal, Poisson, Chi-Square and Cauchy distributions. Permutation of raw data with Euclidean distance method can be recommended when the sample size is less than 15 and permutation of residuals under the full model with Euclidean distance method can be preferred when the sample size is higher than 15 except for normal distribution. Bray-Curtis distances are not suitable for such distributions.