[Paper Review] The Impact of Levene's Test of Equality of Variances on Statistical Theory and Practice

Joseph L. Gastwirth, Yulia R. Gel+1 more|Oct 2, 2010

Advanced Statistical Methods and Models93 references3 citations

TL;DR

This paper reviews Levene's test for equality of variances and its robust modifications, emphasizing its power and resilience to nonnormality. It introduces a trend-adjusted Levene test to detect monotonic variance changes—critical in fields like finance and ecology—demonstrating broad applicability across disciplines such as medical measurement, economics, and environmental science.

ABSTRACT

In many applications, the underlying scientific question concerns whether the variances of $k$ samples are equal. There are a substantial number of tests for this problem. Many of them rely on the assumption of normality and are not robust to its violation. In 1960 Professor Howard Levene proposed a new approach to this problem by applying the $F$-test to the absolute deviations of the observations from their group means. Levene's approach is powerful and robust to nonnormality and became a very popular tool for checking the homogeneity of variances. This paper reviews the original method proposed by Levene and subsequent robust modifications. A modification of Levene-type tests to increase their power to detect monotonic trends in variances is discussed. This procedure is useful when one is concerned with an alternative of increasing or decreasing variability, for example, increasing volatility of stocks prices or "open or closed gramophones" in regression residual analysis. A major section of the paper is devoted to discussion of various scientific problems where Levene-type tests have been used, for example, economic anthropology, accuracy of medical measurements, volatility of the price of oil, studies of the consistency of jury awards in legal cases and the effect of hurricanes on ecological systems.

Motivation & Objective

• To evaluate the robustness and practical utility of Levene’s test for homogeneity of variances in non-normal data.
• To address the limitation of standard Levene tests in detecting monotonic trends in variance, such as increasing volatility in financial data.
• To extend Levene-type tests for use in multi-group, multi-factor, and complex survey designs where variance heterogeneity affects inference.
• To examine the impact of dependence, missing data, and nonresponse on Levene-type test validity in observational studies.
• To stimulate development of sensitivity analyses for unobserved confounders in variance testing, analogous to methods in mean-comparison studies.

Proposed method

• Adapts Levene’s original method by applying the F-test to absolute deviations from group means (or medians) to test for equal variances.
• Proposes a modified Levene test that incorporates trend detection by regressing absolute deviations on group order to identify increasing or decreasing variance patterns.
• Uses large-sample theory and simulation to assess power and robustness under non-normal distributions, including heavy-tailed and skewed data.
• Applies the test to real-world datasets from diverse fields, including medical measurement accuracy, oil price volatility, and jury award consistency.
• Considers the impact of dependence and missing data through simulation and case study analysis, recommending imputation and dependence modeling in practice.
• Extends the framework to complex survey designs by adjusting standard errors and test statistics for stratified, cluster-sampled data.

Experimental results

Research questions

• RQ1 How does Levene’s test perform under non-normal distributions compared to traditional F-tests?
• RQ2 Can Levene-type tests be modified to detect monotonic trends in variance, such as increasing volatility in financial or ecological time series?
• RQ3 What is the effect of dependence and missing data on the Type I error rate and power of Levene-type tests in observational studies?
• RQ4 How do unequal group sizes and many treatments affect the robustness of Levene’s test in multi-factor ANOVA designs?
• RQ5 To what extent can sensitivity analyses for unobserved confounders be adapted to variance testing, similar to methods used in mean-comparison studies?

Key findings

• Levene’s test is robust to non-normality and maintains good power even under skewed or heavy-tailed distributions.
• The modified Levene test with trend detection significantly improves power to detect increasing or decreasing variance patterns, especially in financial and ecological data.
• Dependence among observations can severely distort Type I error rates, necessitating careful experimental design and dependence modeling.
• Missing data, particularly when non-ignorable (e.g., correlated with age or other covariates), can bias Levene-type test results, requiring imputation or sensitivity analysis.
• In complex survey designs, standard Levene tests require variance estimation adjustments to maintain correct size and power.
• The paper demonstrates the test’s utility in diverse applications, including jury award consistency, medical measurement accuracy, and hurricane impact on ecosystems.