Mann-Whitney U Test Calculator

Paste two samples and this returns U for each, the p-value, the ranks behind it and the effect size. For samples of twenty or fewer with no tied values it enumerates the exact null distribution rather than approximating it, and reports how many rank arrangements that was; past that size, or as soon as anything is tied, it switches to the normal approximation with a continuity correction and the tie correction to the variance, and says in the result which of the two it used. A second form gives the p-value and the critical U from sample sizes and a U you already have. The test asks whether one sample tends to give larger values than the other, which is not the same question as a difference in means.