Compare spread across groups using median-centered absolute deviations. Review summaries, p-value, graph, and exports instantly. Useful for audits, research, teaching, and quality checks today.
| Observation | Group A | Group B | Group C |
|---|---|---|---|
| 1 | 12 | 18 | 11 |
| 2 | 14 | 21 | 10 |
| 3 | 15 | 17 | 13 |
| 4 | 13 | 20 | 9 |
| 5 | 16 | 22 | 14 |
| 6 | 12 | 19 | 12 |
This calculator transforms each observation into an absolute deviation from its group center: Zij = |Yij - Ci|.
The center Ci can be the median, mean, or trimmed mean. Median-centered analysis is often preferred when data contain skewness or outliers.
The test statistic is an ANOVA-style F ratio applied to the deviation values:
F = ((N - k) / (k - 1)) × [Σ ni(Z̄i - Z̄)2 / ΣΣ(Zij - Z̄i)2]
The p-value is computed from the F distribution with degrees of freedom (k - 1, N - k).
A small p-value suggests that at least one group differs in variability. A larger p-value suggests the observed variance differences may be explained by sampling noise.
Median-centered analysis is commonly chosen when distributions are not perfectly normal. Mean-centered analysis matches the classic Levene procedure. Trimmed mean can balance robustness and sensitivity.
It checks whether groups have similar spread. The procedure compares absolute deviations from a chosen group center, then evaluates those deviations with an F statistic.
The median-centered form is more robust when data contain outliers or skewness. Many analysts prefer it when normality is doubtful.
Not exactly. The classic form centers observations around the mean. This calculator also supports median and trimmed mean centers for more robust variance testing.
This page provides four input areas. You can compare two, three, or four groups as long as each included group contains at least two observations.
A significant p-value suggests that variability is not equal across all groups. It does not identify which specific group differs most.
No. It is often used because it is less sensitive than raw variance comparisons. The median-centered version is especially useful with imperfect normality.
Yes. The method handles different sample sizes across groups. Just ensure every included group has enough observations for a meaningful spread estimate.
Consider robust methods, variance-stabilizing transformations, or analysis procedures that do not assume equal variance. Your next step depends on study design and goals.