유의확률 분포를 이용한 정규성 검정법의 비교

Abstract

Most statistical inferences and analyses assume that the data are normally distributed. It is thus required to confirm the normality assumption for the corresponding procedures to be valid. The ritual application of statistical procedures without checking the normality assumption may have serious effects on the relevant decisions, for the procedure is based on unjustified theoretical assumption. This study is concerned with comparative studies on the existing test methods of normality assumption which underlies theoretical and empirical studies as well. Most of the normality testing methods have been evaluated using the power of the test. It is, however, well known that the power of the test depends on significance level, set before the experiment, the sample size and the alternative hypothesis. Based on the selected normality testing methods, this study summarizes the factors affecting their performances, and evaluates the methods for the various conditions in terms of their powers and the features of their p-value distributions under the alternatives.