統計學國際知名期刊《Statistica Sinica》近日在線發表了統計學院博士生劉關福的研究論文「Using differential variability to increase the power of the homogeneity test in a two-sample problem」(Statistica Sinica,2016,doi:10.5705/ss.202016.0026;華東師範大學為第一單位)該篇論文合作者為約克大學的 Yuejiao Fu 教授,滑鐵盧大學的 Pengfei Li 教授和華東師範大學的濮曉龍教授。
統計學國際知名期刊《Statistica Sinica》在線發表論文
該研究以醫學實驗和遺傳學中的問題為背景,在兩樣本下首次提出使用EM檢驗方法檢測模型的齊次性。針對EM檢驗方法,該研究給出其在原假設下的極限分布,這個極限分布的形式非常簡單,這在假設檢驗中非常重要,可用於計算檢驗的臨界值。該研究是在位置-尺度分布下討論的,其中,證明參數的似然估計的相合性在目前的文獻中也是第一次,儘管其證明非常複雜,但是此項研究依然給出了嚴謹可靠的證明過程。另外,此研究還向讀者提供了EM檢驗在局部備擇假設下的極限分布,這套理論可用於樣本量的計算。
比較了EM檢驗和現有方法MLRT的功效, EM檢驗的功效在多數情況下都高於MLRT。
EM檢驗(實線)和MLRT(虛線)的功效比較
作者所提出的EM檢驗方法被用於醫學實驗中藥效的檢測和遺傳學疾病的檢測中,其表現效果較好,能夠準確的判斷藥效和疾病的存在,這也再次表明此方法具有優良的特性。
《Statistica Sinica》是國際泛華統計學會會刊,在業內具有極高的影響力,每年一月,四月,七月和十月發刊。
Abstract:We consider a particular two-sample homogeneity testing problem which is often encountered in case-control studies with contaminated controls, or detecting a treatment effect when some subjects are not affected by treatment in biological experiments. To utilize the information from not only the mean shift but also possible change in variance, we propose an EM-test which is designed to simultaneously detect both mean difference and differential variability in the two samples. We show that the EM-test statistic has a chi-square null limiting distribution. The asymptotic properties under local alternatives are also investigated. The main results are established for general location-scale families. Simulation results show that the EM-test possesses more accurate empirical type I error and higher power than the existing methods. Finally, two real data examples are given to illustrate the application of the proposed method.
Keywords:Differential variability, EM-test, Homogeneity test, Limiting distribution, Local power, Mixture models, Two-sample problems.
圖文排版 | 學部信宣部 王坤燕