Abstract

In this article, we propose a family of flexible shrinkage estimators which can use geometric or arithmetic means of sample variances of all genes as the target statistic for shrinkage and also allow the non-normal distribution of the data. The optimal shrinkage parameters are derived under the squared log error loss function and estimated by consistent estimators under a general asymptotic framework. The proposed estimators were evaluated through a series of Monte-Carlo simulation studies and applied to real data from a gene expression study of leukemia patients.

中文翻译：

---

高维基因表达方差的灵活收缩估计器族

摘要

在本文中，我们提出了一系列灵活的收缩估计器，它可以使用所有基因样本方差的几何或算术平均值作为收缩的目标统计量，并且还允许数据的非正态分布。最佳收缩参数是在平方对数误差损失函数下得出的，并在一般渐近框架下由一致的估计器估计。提议的估计量通过一系列蒙特卡罗模拟研究进行了评估，并应用于白血病患者基因表达研究的真实数据。