| In the context of evidential analysis, we investigate the possibility of "robust" likelihood functions for the parameter of interest. Different modes of equivalence of likelihood functions will be introduced first, with respect to which the concept of robustness is illuminated.;Given the knowledge of the distribution that generates the data, the profile likelihood functions are shown to provide asymptotically correct answers for a particular type of evidential analysis. We also define estimated likelihood functions and show that they are asymptotically valid for a different type of analysis. Subsequently, robust likelihoods are introduced, which provide asymptotically correct evidential assessment about the parameter of interest under various models. |
