Outliers may affect the results of classical statistical methods. The method of Analysis of variance (ANOVA) is also not excluded. So the results may led to wrong acceptance or rejection of null hypothesis. In this paper the permutation distribution approach is implemented to find distribution of modified F statistics based on median instead of mean. This approach reduce sensitivity to presence of outliers and classical assumption such as normality. The proposed methods is compared with classical method and robust ANOVA based on forward search. In comparison to competitor method, the introduced approach is free of restricted parametric assumptions. The numerical survey for real and synthetic data demonstrates good performance of proposed method.
