Check whether groups have equal variances before you compare their means.
Levene's test checks one of the assumptions behind the independent-samples t-test and ANOVA: that the groups have equal variances (homogeneity of variance). If one group's scores are much more spread out than another's, the pooled standard error used by those tests is wrong and their p-values can be misleading, especially when group sizes differ.
The test transforms each score into its absolute distance from its group's centre and then runs a one-way ANOVA on those distances. If the groups differ in average distance, they differ in spread. A significant result (p < .05) means the equal-variances assumption is violated.
StatRyx uses the median-centred version, also called the Brown–Forsythe test, which is robust when the data are not normally distributed. The result gives F with both degrees of freedom, the p-value, and the variance and standard deviation of each group, followed by a plain recommendation: when variances are equal, the standard pooled tests are appropriate; when they are not, a Welch t-test or Welch ANOVA with Games–Howell post-hoc tests is the safer choice.
Levene's test is a check, not an end in itself. With very large samples it flags trivial differences, and with small samples it can miss real ones — so look at the group standard deviations as well as the p-value.
| Group | n | SD | Variance |
|---|---|---|---|
| Lecture | 30 | 5.29 | 28.01 |
| Seminar | 30 | 7.21 | 52.03 |
| Online | 30 | 10.21 | 104.23 |
| Levene (median) | F(2, 87) = 7.29 | p = .001 |
Levene's test indicated that the variances of exam scores differed across the three teaching methods, F(2, 87) = 7.29, p = .001 (SDs: lecture = 5.29, seminar = 7.21, online = 10.21). Group means were therefore compared with Welch's ANOVA.
Yes. You can run it without an account (3 free analyses), and a free account adds 10 more — a one-time allowance, no payment needed. Every result includes the APA 7 write-up. Paid plans (Student from $9/month) add more analyses and Word, PDF and Excel export.
Give F with both degrees of freedom and p: "Levene's test indicated unequal variances, F(2, 87) = 7.29, p = .001." Then say which test you used because of it, for example Welch's ANOVA.
That the group variances are not equal, so the equal-variances assumption of the standard t-test or ANOVA is violated. Use Welch's t-test or Welch's ANOVA, which do not assume equal variances.
Bartlett's test is more powerful when the data are normal but is very sensitive to non-normality. Levene's test, especially the median-centred Brown–Forsythe version StatRyx uses, stays accurate when the data are skewed.
When you run a t-test or ANOVA, StatRyx checks the variance assumption as part of the analysis. Ask for Levene's test directly when you want the full table to report.
Run a Levene's Test on your data →