Levene's Test Calculator

In short Levene's test checks whether several groups have similar SPREAD, which is one of the three assumptions behind a standard ANOVA and the one most often violated in real data. The three groups loaded above have variances of 0.700, 38.500 and 2.500 — a 55-fold range — and give F = 4.9299 on 2 and 12 degrees of freedom with p = 0.0274, so the equal-variance assumption fails and a standard ANOVA on that data would be the wrong test. The mechanism is neat and worth knowing: replace every value with its absolute distance from its own group's MEDIAN, then run an ordinary one-way ANOVA on those distances. If the groups have similar spread, the distances are similar too and F is small. The test is an ANOVA in disguise. Using the median rather than the mean is the Brown-Forsythe variant, and it is what makes this robust — the original mean-based version collapses on skewed data in much the way the F test for variances does. If the test fails, the answer is not to abandon the analysis: use Welch's ANOVA for means, or Kruskal-Wallis for distributions. And read the variance ratio as well as the p-value, because with small groups this test has little power to detect a difference that is plainly there.