Empirical likelihood is a distribution-free approach that allows one to construct likelihood functions without knowing the true underlying distribution. Modification has been proposed to ensure that the large sample property is better achieved when sample size is not large or when there are many parameters. Alternatively, one can employ the parametric robust likelihood procedure proposed by Royall and Tsou (2003) to make likelihood inference under model misspecification. We give a thorough comparison between the two model-independent robust likelihood approaches and show that the method by Royall and Tsou (2003) is superior to the empirical likelihood in terms of various performance benchmarks.