need help

What are the necessary conditions on the distributions of the data in a CRD in order for the Kruskal- Wallis statistic to be valid in testing for treatment differences? How are these conditions different from the conditions required for applying the F = MSTRT/MSE test statistic? The ANOVA from a completely randomized design with r=10 reps of 5 treatments yielded MSError = 9. There were N = 50 observations in the experiment with the residuals ranging from -6.5 to 8.7. Would you consider any of the residuals to be outliers? Justify your