本文主要的目的是試著利用推廣一般反應變數獨立時的估計函數，加入以處理非獨立反應變數為目的的不偏估計函數，來探討這種複合估計函數在有效性上的表現。 研究中將納入多個具有不偏性的複合估計函數，來比較其優劣。 The work of this thesis is inspired by an observation of the score function of the multivariate negative binominal distribution. We propose composite estimating equations consisting of two parts: unbiased estimating equations for independent responses and unbiased estimating equations taking into account intra-cluster dependence. Simulation studies are launched to compare a number of composite estimating equations in terms of efficiency for regression parameter estimations.
