本論文首先介紹Royall與Tsou在2003年所提出之強韌概似函數的方法。其次，將其方法與觀念，應用於分析相關性的有序(ordinal)資料並於資料平均數在廣義線性模型的架構下，推導大樣本時，有興趣之迴歸參數的強韌概似函數。最後，將平均數由廣義線性模型進ㄧ步推廣到部分線性模型的架構且同樣推導大樣本時，有興趣之迴歸參數的強韌概似函數。 值得注意的是這些概似函數並不需要知道資料的真實分配，只需要假設二階或四階動差存在即可。最後，利用模擬與真實資料的分析來呈現此強韌方法的效率。 In this thesis, we first introduce the idea of robust likelihood functions proposed by Royall and Tsou (2003). Next, we provide a parametric robust method originated from this idea to make inferences for correlated ordinal data and develop the robust likelihood functions for regression coefficients of mean modeled in a generalized linear model fashion. Finally, we extend the robust likelihood technique from generalized linear models (GLM) to partially-linear models (PLM), and use normal distribution as the working model to develop the robust likelihood functions for regression coefficients in large samples. The legitimacy of this novel approach requires no knowledge of the underlying joint distributions so long as their second or fourth moments exist. The efficacy of the proposed parametric approach is demonstrated via simulations and the analyses of several real data sets.
