486 – Computing Kaleidoscope
On Consistency and Limitation of Parametric and Non-Parametric Paired Sample Tests
Khairul Islam
Department of Mathematics and Statistics, Eastern Michigan University
Tanweer Shapla
Department of Mathematics and Statistics, Eastern Michigan University
A paired t-test is a very common test for before-after measurements under the assumption that the difference of the before-after measurements form a sample form a normal distribution. In the violation of normality assumption, the popular alternatives are nonparametric sign test or signed rank test. How do these tests behave in the presence of varying skewness or varying correlation in the paired population? In this paper, we carry out a simulation study to evaluate the consistency and limitations of various tests of paired samples due to varying skewness and correlation in the paired populations from beta, exponential and gamma distributions.