Generation of Non-normalData: A Study ofFleishman’s PowerMethod
2011 (English)Report (Other academic)
Fleishman's power method is one of the traditional methods used for generatingnon-normal random numbers. In this paper, we use Monte Carlo simulationto test the reliability of this method. Specially, we assess the performance of the method under different conditions of skewness, kurtosis and sample sizes.The power of the normality test statistics proposed by D'Agostino (1986) isstudied based on the generated samples. The results suggest that Fleishman'smethod has difficulties on generating non-normal samples with higher levels of skewness and kurtosis. The effect of sample size is found to be significant on the reliability of the data generation. The parabola, which indicates the bottomboundary of the possible combination of skewness and kurtosis calculated by Fleishman (1978), is shown to be incorrect. When it comes to the power study, a considerable impact of sample size is also observed on obtaining a trustworthytest decision based on the generated non-normal samples.
Place, publisher, year, edition, pages
Uppsala: Department of Statistics, Uppsala University , 2011. , 29 p.
Working paper / Department of Statistics, Uppsala University, 2011:1
non-normal data, Fleishman's method, D'Agostino test of normality, power of the test
Probability Theory and Statistics
IdentifiersURN: urn:nbn:se:uu:diva-150623OAI: oai:DiVA.org:uu-150623DiVA: diva2:407995