不能T检验，就非参数检验啊

To respond,

1. Non-parametric test is qualified to compare median of two independent samples, rather than mean. The statistical power of non-parametric is relatively under expected, however, could provide you unbiased estimation, particularly when the assumption of normality is violated.

2. The transformation is always risky even using identical mapping method. It is well developed for certain distribution so far. However, when you are unable to do so, go for non-parametric approach in case of increasing uncertainty.