Abstract
Normality is one of the main presuppositions in statistical tests. The multiplicity of the normality tests bring out another problem of choosing the appropriate test for researchers. The free software R which has a great popularity in the statistical analysis has 18 normality tests in 4 different packages. In this study we compared performance of these normality tests in terms of empirical type 1 error rate and power by Monte Carlo simulation. As a result, regardless of the distribution of data (symetric or asymmetric) the Shapiro-Francia test, also the Frosini B test performed better than the other normality tests in terms of experimental type 1 error rate. However the widely used Kolmogorov-Smirnov test showed worse performance than other normality tests in terms of empirical type 1 error rate and power.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
