Classical parametric tests such as t-test is a powerful analytical tool to test the equality of central tendency for two groups. However, the use of parametric tests restricted to certain assumptions such as the data must be continuous and normally distributed, variances for different groups must be homogeneous, data must be randomly sampled and the observations must be independent. When the situations do not meet these assumptions, especially when the data is continuous but not normally distributed or with the small sample size, the Type I error and power rates will be affected drastically. Therefore, non-parametric test is an alternative for researcher when normality assumption of parametric test is violated. However, due to loss of information when using non-parametric test, researchers tend to find more alternative tests such as resampling methods. Bootstrapping and permutation test are the focus in this study. Performance of parametric tests, non-parametric tests, bootstrapping and permutation test in different simulated situations were measured in term of Type I error and power of test and compared via Monte Carlo studies. Through this study, permutation test perform better than bootstrapping in most cases. Overall, resampling methods perform better than parametric test when normality assumptions are violated.