Abstract: For linear quantile regression model, this paper proves that the test statistics, besed on smoothed empirical likelihood (SEL) method and least absolute deviation (LAD) method, both converge weakly to a noncentral Chi-square distribution under the local alternatives H_1:\beta=\beta_0+a_n, where \beta is the true parameter. Simulation results show that the SEL method is more efficient than the LAD method.