Year: 2011
Type: article
Abstract: The purpose of this study was to compare the efficiency of commercial statistical packages in testing normality.The six (6) tests being studies were Kolmogorov, Lilliefors and Shapiro-Wilk available in SPSS14 and Kolmogorov, Anderson-Darling and Ryan-Joiner available in MINITAB 14.The data for this study was obtained from simulation, by the method of Monte Carlo, under conditions of a normal distribution and slightly different from the normal distribution, by using Calc-Random Data menu of MINITAB 14.In each situation, 500 iterations were carried out with different sample size: 10, 20, 30, 50 and 100.Comparison of type I error rate and empirical power among the six test statistics were also made.It was found that Ryan-Joiner Test in MINITAB 14 had the highest empirical power in all cases and sample sizes, at a significance level of 0.10.The sample size of 100, in particular, showed the empirical power at almost 1.It also had an ability to control probability of Type I error in almost all situations under the criteria of Cochran.Nonetheless, the K-S test in MINITAB 14, appearing under the appellation of the K-S test was in fact the Lilliefors test.
Source: Science & Technology Asia
Author: Umaporn Chantasorn
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W2403463548
