空间动态面板模型拟极大似然估计的渐近效率改进.pdfVIP

空间动态面板模型拟极大似然估计的渐近效率改进 ∀ 145 ∀ 空间动态面板模型拟极大似然 估计的渐近效率改进 1 2 张征宇 朱平芳 ( 1 上海财经大学经济学院; 2 上海社会科学院数量经济研究中心) ! Lee 和Yu ( 2008) 研究了一类同时带 个 与时间固定效应的空间动态 面板模型的拟极大似然估计量的大样本性质本文说明当扰动项非正态时, 拟极大 似然估计量的渐近效率可以被进一步提高为此, 我们构造了一组含待定矩阵且形 式一般的矩条件以用来包含对数似然函数一阶条件的特殊形式从无冗余矩条件的 角度, 选取最优待定矩阵得到了最佳广义矩估计量本文证明了当扰动项正态分布 时, 最佳广义矩估计量和拟极大似然估计量渐近等价; 当扰动项非正态分布时, 广 义矩估计量具有比拟极大似然估计量更高的渐近效率Monte Carlo 实验结果与本 文的理论预期一致 空间面板模型 拟极大似然估计 广义矩估计 渐近效率 F2240 A Improvement on Asymptotic Efficiency of QMLE for SDPD Model Abstract: Lee and Yu ( 2008) investigates the asym totic ro erties of quasi maximum likelihood estimator ( QM LE) for a class of s atial dynamic anel data ( SDPD) models in the resence of both individual and timew ise fixed effectsWe show in this a er that the asym totic efficiency of QM LE can be further im roved under nonnormalityWe carefully construct one set of moment conditions that are general enough to include the s ecific form of the first order conditions of the log likelihood function as a s ecial caseT he best generalized method of moments esti mator ( BGM ME) in terms of nonredundant moments is derivedThe resulting GM ME is show n to be asym totically equivalent to QM LE under normality and w ill be more efficient when the disturbances are not normally distributedMonte Carlo results seem to be consistent w ith our theories Key words: SDPD models; QMLE; GM M; Asym totic Efficiency 2008 # ∃ , ∀ 146 ∀ #数量经济技术经济研究∃ 2009 年第5 期 ( S atial econometrics) , ( s atial effects) , , ( s atial autocorrelations) ( Anselin Bera ( 1998) ) ( A nselin, Le Gallo