| Summary: | 碩士 === 國立成功大學 === 數學系應用數學碩博士班 === 91 === Houng (1998) has shown that in nonlinear regression models with means expressed as , where can also be a vector, and under the composite hypothesis with respect to the alternative , the score test statistic has the robustness property, i.e., it is independent of the form of f. In this article, we will also under the same models, when there is some assumed relationship between the mean and variance of each observation but not necessarily a fully specified likelihood. We can deal with this kind of problems by using quasi-score test statistics, and it will be proved that the quasi-score test statistics are still robust. Finally, simulations and a practical example will also be presented to illustrate these results, and compare with score test statistics.
|
|---|
