A new procedure to analyze the results of experimental designs is intr
oduced. A permutation procedure based on Euclidean distance is used to
evaluate residuals obtained from least sum of absolute deviations reg
ression. Applications include completely randomized and randomized blo
ck configurations including one-way, factorial, split-plot, and Latin
square designs, with or without covariates. Parametric assumptions of
homogeneity of variance, compound symmetry, and normality are eliminat
ed with this procedure.