结构方程模型的信息损失

Information Loss in Structural Equation Models

利用统计曲率的概念 ,研究结构方程模型的最大似然估计量和广义最小二乘估计量的信息损失 ,得到了简明的结果 .证明了该模型的信息损失仅依赖于固有曲率 ,而与参数效应曲率无关 .另外 。

By employing the concept of statistical curvatures,the information loss of the maximum likelihood estimator and the generalized least squares estimator is investigated.Simple expressions are obtained.It is shown that the information loss depends on the intrinsic curvature only,while the parameter effect curvature has no effect at all.Moreover,the information loss of theg eneralized least squares estimator depends seriously on the choice of the weighting matrix.