The study compared seven statistics for testing the difference between two groups when data didn't meet assumptions. The Welch Based on Rank test was best for log-normal, gamma, and poisson distributions with equal variance. The Welch t test performed well for various distributions with unequal variance.