Levene's Test

What Is Levene's Test?

Levene's test is a statistical method used to check whether two or more groups have equal variances — a property known as homogeneity of variance (or homoscedasticity). Variance is a measure of how spread out scores are around the group mean. If one group's scores are tightly clustered while another group's scores are widely scattered, the variances are unequal, and several common statistical tests may produce misleading results. Levene's test gives you a formal way to check this assumption before proceeding with your main analysis.

Why Do We Need It?

Many of the most popular statistical tests — including the independent samples t-test and the one-way analysis of variance (ANOVA) — assume that the groups being compared have roughly equal variances. When this assumption is violated, the p-values produced by these tests can be too small or too large, leading you to either find effects that are not really there or miss effects that are.

Imagine a researcher comparing exam scores between two schools. School A has a mean of 72 with scores ranging from 60 to 84, while School B has a mean of 68 with scores ranging from 30 to 100. Even though both means are close, the spread of scores is dramatically different. Running a standard t-test without checking this difference in spread could yield an unreliable conclusion. Levene's test alerts you to the problem so you can take corrective action — for instance, by switching to Welch's t-test, which does not assume equal variances.

How Does It Work?

The core idea behind Levene's test is elegant. For each observation, the test calculates the absolute difference (or, in some variants, the squared difference) between that observation and the mean of its group. These absolute deviations become a new set of scores. Levene's test then performs a standard one-way ANOVA on these deviation scores. If the groups have equal variances, their average deviations from the group means should be similar, and the ANOVA will yield a non-significant result. If one group is much more spread out than the others, its deviations will be larger, the ANOVA will detect this, and the result will be significant.

A common variant replaces the group mean with the group median when computing the deviations. This median-based version (sometimes called the Brown-Forsythe test) is more robust to non-normal data and is the version used by many statistical software packages by default. The test produces an F-statistic and an associated p-value, just like a regular ANOVA.

What Does the Result Mean?

Interpreting Levene's test requires a small mental flip compared to most hypothesis tests. Here, a non-significant result is the outcome you usually hope for. A non-significant p-value (typically above 0.05) means there is not enough evidence to conclude that the variances differ — in other words, the equal-variance assumption appears to hold, and you can proceed confidently with your t-test or ANOVA.

A significant result (p-value below 0.05) indicates that the group variances are significantly different. This does not invalidate your research question — it simply means you should use a statistical method that does not require equal variances. For a two-group comparison, Welch's t-test is the standard alternative. For three or more groups, you can use Welch's ANOVA or the Brown-Forsythe ANOVA, both of which adjust for unequal variances. Non-parametric tests, which do not assume equal variances, are another option.

Key Assumptions

• Independent observations: The scores within and between groups must be independent. One participant's score should not influence another's.
• Continuous data: The outcome variable should be measured on a continuous (interval or ratio) scale, since variance is only meaningful for continuous data.
• Random sampling: The data should come from randomly selected or randomly assigned samples to ensure the test results generalise to the broader population.

Importantly, Levene's test is relatively robust to departures from normality, especially when the median-based variant is used. This is one reason it is preferred over older tests of equal variance, such as Bartlett's test, which is highly sensitive to non-normality.

When to Use It

Run Levene's test as a preliminary check before any analysis that assumes equal variances. The most common situations are before an independent samples t-test or a one-way ANOVA. It is also useful in more complex designs (such as factorial ANOVA) where the homogeneity of variance assumption applies. Keep in mind that with very large samples, Levene's test can flag trivially small differences in variance as significant, so it is good practice to also inspect the variances visually (for example, with boxplots) and use your judgement about whether the difference is large enough to matter in practice.

A Quick Example

A sociologist compares income satisfaction scores across three occupational groups, each with 25 participants. Before running a one-way ANOVA, she performs Levene's test and obtains F = 4.83 with a p-value of 0.011. Because 0.011 is below 0.05, she concludes that the variances across the three groups are significantly different. Rather than proceeding with a standard ANOVA, she switches to Welch's ANOVA, which does not require equal variances, and reports her findings accordingly.

Levene's test is a small but critical step in responsible data analysis. By checking the equal-variance assumption before your main test, you protect the validity of your conclusions and demonstrate careful, rigorous statistical practice.