Abstract:
|
A parametric two-sample t-test is the most powerful test for comparing two population means under normal models. In real-life, the assumption of normality may not meet. Under these circumstances, a transformed two-sample t-test or alternately, analogous non-parametric tests such as Mann-Whitney test and Kolmogorov Smirnov tests may be employed to achieve certain objectives. To recommend the best test under non-normal models, it is imperative to evaluate performance of underlying tests using Type I error probability and power of the test via Monte Carlo simulation at various sample sizes. In this study, we simulate independent samples from skewed distributions with varying levels of skewness, along with symmetric distributions to evaluate performance of four underlying tests, namely, two-sample t-test, transformed two-sample test, Mann-Whitney test and Kolmogorov Smirnov test.
|