Small-Sample Bias in GMM Estimation of Covariance Structures
研究协方差结构模型中广义矩估计（GMD）的小样本性质，发现最优最小距离估计量（OMD）存在向下偏差，且通常劣于等权最小距离估计量（EWMD）；同时提出一种无偏替代估计量，但蒙特卡洛证据显示其仍不如EWMD。
We examine the small-sample properties of the generalized method of moments estimator applied to models of covariance structures, in which case it is commonly known as the optimal minimum distance (OMD) estimator. We find that OMD is almost always biased downward in absolute value. The bias arises because sampling errors in the second moments are correlated with sampling errors in the weighting matrix used by OMD. Furthermore, OMD is usually dominated by equally weighted minimum distance (EWMD). We also propose an alternative estimator that is unbiased and asymptotically equivalent to OMD. The Monte Carlo evidence indicates, however, that it is usually dominated by EWMD.