when working with independent non-Gaussian distributions, I would use the Mann-Whitney test to compare two independent samples (similar to the unpaired t test for normal distributions) or the Kruskal-Wallis test to compare three or more samples within each other (similar to one-way Anova for normally distributed samples).

Inspect the histograms and QQ-plots. The t-test is known to be quite robust. In doubt, nonparametric and resampling alternatives to t-tests are available.

Nonparametric (as suggested by Paulo):

https://www.statmethods.net/stats/nonparametric.html

Resampling:

https://www.statmethods.net/stats/resampling.html

Hi Dimas Wicaksono

Despite the fact that your n=15 is low, the tests mentioned in your questions depends on the number of groups. Since you have stated that your data is non-parametric, you would want to use Mann-Whitney if you want to compare medians of two groups (whether median of one group is larger than the other) or Kruskal-Walis if you want to compare three or more groups. Nonetheless, t-test is robust enough even with nonparametric data as rightly stated by Jos Feys . My problem though is your small sample size. It would also help you you indicated which groups (how many groups) you're comparing.

Best,

Mangats ya, Dim