Mehmet Mendes
Department of Genetics, Faculty of Agriculture Biometry,
Akin Pala
Department of Animal Science, Canakkale Onsekiz Mark University, Canakkale, Turkey
ABSTRACT
In this study, Shapiro-Wilks, Lilliefors and Kolmogorov-Smirnov tests were compared for Type I error and for power of the tests. The simulation was run 100, 000 times for different situations and for different types of departures from normality. For all different sample sizes and distributions, Shapiro-Wilks gave the most powerful results, followed by the Lilliefors test. Kolmogorov-Smirnov test results were the weakest among all three tests. All three test were most powerful when ran on data with exponential distribution.
PDF References Citation
How to cite this article
Mehmet Mendes and Akin Pala, 2003. Type I Error Rate and Power of Three Normality Tests. Information Technology Journal, 2: 135-139.
DOI: 10.3923/itj.2003.135.139
URL: https://scialert.net/abstract/?doi=itj.2003.135.139
DOI: 10.3923/itj.2003.135.139
URL: https://scialert.net/abstract/?doi=itj.2003.135.139
REFERENCES
- Shapiro, S.S. and M.B. Wilk, 1965. An analysis of variance test for normality (Complete samples). Biometrika, 52: 591-611.
CrossRefDirect Link