Selection of Statistical Significance Test according to the Distribution of the Visual Evaluation in Diagnostic Imaging

Abstract

We examined how to select the method of statistical significance test as to the evaluation obtained by diagnostic imaging.We used 6-10 pairs of small data, and classified the distribution style using skewness and kurtosis.In statistical significance test, we used the paired t-test as a parametric test, and the Wilcoxon test as a non-parametric test.We tried to select the method of the statistical significance test in consideration of the power of test in each distribution style.The result of the X^2test, Which is a test of normalization, there was no significance.In the case the paired t-test may be adecuate, considering the power of test, we should use the Wilcoxon test, incase that the both bottoms spread or the right bottom spreads.