Comparison of normality tests with different sample sizes

One of the most important assumption for parametric tests is normality of a distribution. Many normality tests are available in the literature. In our study we compare 5 normality tests which are most popular and available in statistical softwares. Usage of normality tests differs due to samples size or nature of data. For that purpose distributions are generated from different theoretical distributions and sample sizes by Monte-Carlo simulation. Type-I error and power used for comparison of normality tests. According to simulation results, Shapiro-Wilk test has the best results, when the sample size decreases, Anderson-Darling has also good results as Shapiro-Wilk test. All normality tests' power getting lower with a decrease of sample sizes. At that situation we suggest to provide normality test results with graphical techniques. Keywords: Normal distribution, Normality tests, Type-I error, Power, Monte-Carlo Simulation