摘要
设m维随机变量X=(X1,X2,…,Xm)的copula函数为C(u1,u2,…,um);α)=C((F1(x1),F2(x2),…,Fm(xm));α),本文在(X1,X2,…,Xm)的样本空间和(U1,U2,…Um)的样本空间上讨论了m元copula函数中参数α的极大似然估计,得到了边缘分布函数连续时,两样本空间上参数α的极大似然估计和最大后验估计的等价性;而边缘分布函数不连续时,两样本空间上参数α的极大似然估计和最大后验估计的渐近等价性.
Let C( ( u1, u2 ,…, um ) ; a ) = C( ( F1 ( x1 ), F2 ( x2 ) ,…, Fm ( xm ) ) ; a ) be the copula of m-dimension random Variables X = ( X1, X2,…, Xm ). The paper discusses Maximum Likelihood Estimation (MLE) of Parmeter on Copula, based on the sample space of X = ( X1 , X2 ,…, Xm ) and the sample space of U = ( U1 , U2,…, Um ). It derives the equivalence of MLE of parameter on the two sample space with continuously marginal distribution function, So is Maximum Posterior Estimation (MPE). However, for discontinuously marginal distribution funtion, MLE and MPE of the parameter on the two sample space are asymptotically equivalent with large sample size.
出处
《经济数学》
2008年第2期210-215,共6页
Journal of Quantitative Economics
基金
国家自然科学基金资助项目(No.10301011)
关键词
COPULA函数
样本空间
极大似然估计
最大后验估计
Copula function, sample space, maximum likelihood estimation, maximum posterior estimation