In applied research, Levene’s test is commonly used to assess the equality of variances among k-independent normal distributions. However, simulation results indicate that both Levene’s test and its modified version tend to be conservative, showing smaller Type I error rates compared to the nominal level for small sample sizes. On the other hand, Owen’s empirical likelihood ratio test (ELRT) provides a robust non-parametric alternative for testing the equality of means in normal distributions, especially with three or more groups. In this paper, an Empirical Likelihood (EL) Version of the modified Levene’s test is proposed. Simulation results suggest that the proposed test is liberal for small sample sizes while maintaining significance levels for moderately large sample sizes. Particularly for moderately large sample sizes, the proposed test exhibits greater power than the modified Levene’s test. Emerson suggests a preference for tests slightly more liberal than conservative ones. The proposed test proves robust against violations of the normality assumption when samples are drawn from symmetric and asymmetric distributions, including log-normal, student’s t, gamma, exponential and chi-square distributions. Despite the EL version of the modified Levene’s test being computationally more intensive than modified Levene’s test, an R program developed by the author makes it easy to use.