Porovnání testů normality na problémových datech různého typu

Abstract

The theme of this thesis is Comparison of Normality Tests on the Problem Data of Different Types. The goal of the bachelor thesis in theoretical part is to provide an of overview the most frequently used method of verifying the normality. Then the graphical methods verification of normality are described. Next part describes tests based on moments, regression testing and goodness of fit tests. The practical part is based on verification of normality on randomly generated data from different distributions. The goal is to verify the normality of the various methods and tests by visual inspection of graphical methods, or by other tests comparing the p-value of the significance level of α = 0.05. Assumption null hypothesis, when verifying the normality, indicates that the assessed data constitute a random selection from a normal distribution. We believe that the assumption of normality is completed, when the null hypothesis is rejected. Testing are performed in statistical software. Next is comparison of the effectiveness of the methods.