<P>GMM with optimal minimum distance(OMD) is biased in small samples when used to estimate covariance models. Equal weighting matrix outperforms OMD but it is inefficient.</P>
<P>各位有什么建议？</P>
<P>看到2、3篇文章想从information theory解决这个问题，比如kullback/leibler information criterion,但是都不很成功。</P>
<P>想法到是有点，但是这个方向值得研究么？</P>
<P>恳请指点，谢谢！</P>