多重假设检验中参数估计问题研究

Estimating the Number of True Null Hypotheses and Its Application in Multiple Testing

在多重假设检验中,真正原假设的个数m_0是未知的,但是它有着很重要的影响,因此,它在最近的统计文献中备受关注。文章综述了三种主要的估计方法:最低斜率法、三次样条法、均值估计方法。然后将上述三种方法结合起来,提出了新的估计方法:均值三次样条法,并主要研究了其在微阵列数据上的应用。大量的模拟研究表明,和其他方法相比,新的估计方法具有较小的偏差和标准差。最后利用真实数据来对估计方法进行评估,并找出了差异表达性基因。模拟和实际数据表明此方法具有显著性提高。

In the multiple tests, the m0 is unknown, and it has an important effect, so it has attracted much attention in the recent statistical literature. This paper reviews three main methods of estimation, Lowest Slope （LSL） method, average estimate approach. Then the new method is proposed by the above methods. We mainly focus on the application of the microarray data in this paper. Extensive simulation studies indicate that this estimate approach has relatively small mean errors and comparing it with the above methods. Finally carry out the mieroarray data to evaluate the performance of the proposed estimator, and find the significant genes. Both simulated and real data can demonstrate that our method may improve the existing literature significantly.