Linear mixed-effects models are widely used in analyzing longitudinal data, where the inference of unknown parameters is of most interest. Although the statistical inference of the fixed effects is well studied, the analysis of variance component under the non-parametric setting still needs more exploration. In this paper, we propose empirical likelihood-based methods for the inference of variance component with known and unknown fixed effects. A non-parametric version of the Wilks’ theorem for the proposed statistics is derived. Simulation studies demonstrate that the proposed methods exhibit better coverage than the commonly used likelihood ratio method under the Gaussian assumption and the results with unknown fixed effects are as good as those with known fixed effects. The new tests are illustrated in the analysis of a wearable device dataset.