Study of Data Transformation Methods to Achieve Normality

AUTHOR(S):

TITLE

A Comparison Study of Data Transformation Methods to Achieve Normality

ABSTRACT

Normality is the one of main important central assumptions in statistical studies. Since in reality this is not the fact, transformation of random variables are required to achieve specified purposes i.e. stability of variance, the additivity of effects and the symmetry of the density. In this study, we make a comparison study in order to check the power of the transformations method for satisfying the normality. We simulated Log-normal, Beta and Gamma probability distributions with various parameters in order to transform them to be normal. The statistical hypothesis tests that are well known to be powerful are used in order to examine the performance of the transformation methods.

KEYWORDS

Data transformations, Monte Carlo simulation, Box-Cox transformation, Normality comparison

REFERENCES

[1] Osborne, Jason, Notes on the use of data transformations, Practical Assessment, Research & Evaluation, Vol. 8, No.6, 2002, Retrieved March 21, 2009. [1] Osborne, Jason, Notes on the use of data transformations, Practical Assessment, Research & Evaluation, Vol. 8, No.6, 2002, Retrieved March 21, 2009.

[2] Hoyle, M.H., Transformations: an introduction and a bibliography, International Statistical Review, Vol.41, No.2, 1973, pp. 203–223.

[3] Tan W.D., Gan F.F., Chang T.C., Using normal quantile plot to select an appropriate transformation to achieve normality, Computational Statistics & Data Analysis, Vol.45, 2004, pp.609 – 619.

[4] Box, G.E.P., Cox, D.R., An analysis of transformations, J. Roy. Statist. Soc. Ser. B, Vol.26, 1964, pp.211-252.

[5] Box, G.E.P. and Cox, D.R., An analysis of transformations revisited, rebutted. Journal of the American Statistical Association, Vol.77, 1982, pp.209–210.

Cite this paper

Ozer Ozdemir. (2016) A Comparison Study of Data Transformation Methods to Achieve Normality. International Journal of Mathematical and Computational Methods, 1, 382-383