考虑线性混合效应模型中随机效应的选择问题. 在随机效应和误差项没有分布假定的条件下, 研究基于修正Cholesky分解的硬阈值估计和一种罚估计在选择随机效应和估计方差分量时的表现. 从理论上证明了这两种估计方法具有相合性, 并且罚估计方法具有神谕性质.

It is important to select random effects in linear mixed effect models. When the distributions of random effects and error terms are unknown, we use Cholesky decomposition to study a hard thresholding estimator and a penalized estimator for selecting random effects in LME model. It is proved that the two kinds of estimators are consistent and the penalized estimator has Oracle property.