Robustness of likelihood specification
Yudi Pawitan · 2001
Abstract In Section 9.4 we have shown that, under general conditions, the likelihood approach leads to the best possible inference. The notion of ‘best ‘, e.g. achieving the Cram´er-Rao lower bound, is measured within a particular model. The likelihood approach requires a full specification of the probability structure, and it is difficult to quantify the uncertainty associated with model selection. The question we will address is how sensitive the result is to the correctness of an assumed model.