偏差改正的Partial EIV模型方差分量估计

针对Partial EIV模型的方差分量估计中未考虑参数估值偏差所带来的影响,将Partial EIV模型视为非线性函数得到参数估值的偏差及二阶近似协方差表达式,计算得到偏差改正后的参数估值,结合方差分量估计方法,更新由参数估值影响的矩阵变量,给出了基于偏差改正的方差分量估计迭代方法。试验结果表明,参数估值及其协方差主要受参数估值偏差大小的影响,加入偏差改正能够得到更加合理的参数估值及方差分量估值,偏差改正后的方差分量估值可更加合理地评估参数估值的精度信息。

Bias-corrected variance components estimation of Partial EIV model

Considering the methods of variance components estimation (VCE) in Partial errors-in-variables (Partial EIV) model have not considered the effect of the bias of parameter estimates, the formulas of bias and second-order covariance matrix of parameter estimates are presented with the Partial EIV model regarded as a non-linear function and the parameter estimates after bias-correct are calculated. Combining the VCE method, the matrix variable influenced by the parameter estimates is updated, and an iterative method of variance components estimation based on bias-correct is given. The experiments show that the reasonable parameter estimates and its second-order approximate covariance results are affected by the bias of parameter estimates. The reasonable parameter estimates and variance components estimates can be obtained through the bias-correct and the second-order information obtained can reasonably evaluate precision of parameter estimates.