含缺失协变量的混合效应模型的多元单边检验  

Multivariate One-sided Tests for Mixed Effects Models with Missing Covariates

作　　者：齐培艳 仵旭靓 段西发 QI Peiyan;WU Xujing;DUAN Xifa(College of Applied Sciences,Taiyuan University of Science and Technology,Taiyuan,030024)

机构地区：[1]太原科技大学应用科学学院,山西太原030024

出　　处：《山西大学学报（自然科学版）》2022年第5期1195-1205,共11页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(11701406);山西省基础研究计划项目(202103021224274);山西省自然基金(20210302123224)。

摘　　要：针对具有缺失协变量的半参数非线性混合效应(SNLME)模型的多元单边检验问题,提出了基于多重插补法的似然比检验(LRT)和Wald检验,并推导了其渐近零分布,将其应用到抗HIV研究中。实例分析表明本文给出的检验统计量远大于基于完全情况(CC)下的值,为病人抗HIV治疗的有效性提供了更有力的证据。The likelihood ratio test(LRT) and Wald test based on multiple imputation are proposed for the multivariate one-side test of semi-parametric nonlinear mixed effect model(SNLME) model with missing covariates, and its asymptotic zero distribution is derived and applied to anti-HIV research. The case study shows that the test statistics are much larger than those based on complete case(CC), which provides more powerful evidence for the effectiveness of anti-HIV treatment.

关 键 词：半参数非线性混合效应模型 多元单边检验 多重插补 似然比检验 WALD检验 

分 类 号：O212.1[理学—概率论与数理统计]