An Exhaustive Power Comparison of Normality Tests

A goodness-of-fit test is a frequently used modern statistics tool. However, it still unclear what the most reliable approach to check assumptions about data set normality. particular (especially with small number of observations) only partly describes process, which leaves many options for interpretation its true distribution. As consequence, statistical tests have been developed, power depends on circumstances (i.e., sample size, outlets, etc.). With aim developing more universal test, we propose an based N-metric our chosen kernel function. To compare 40 normality tests, hypothesis was tested 15 distributions 6 different sizes. Based exhaustive comparative research results, recommend use samples size n≥118.