Ecologists tested different statistical methods on ecological data and found that ANOVA and randomization tests are robust to non-normality, uneven sample size, and unequal variance. However, the t-test for unequal variance lost power with uneven sample sizes. Non-parametric tests were affected by unequal variance in large samples but could complement parametric tests for uneven sample sizes. The combination of parameters like distribution, sample size, and equality of variances strongly influences the performance of statistical tests. Manuals should provide more detailed instructions on when to use specific tests.