无金标准部分核实数据下基于Bootstrap的小样本齐性检验  

Homogeneity Tests Based on Bootstrap Method for Small Sample Size of Partially Validated Data Without Gold Standard

作　　者：伏启翔 覃愿 王黎明 邱世芳 FU Qixiang;QIN Yuan;WANG Liming;QIU Shifang(College of Science,Chongqing University of Technology,Chongqing 400054,China)

机构地区：[1]重庆理工大学理学院,重庆400054

出　　处：《重庆理工大学学报（自然科学）》2021年第2期289-300,共12页Journal of Chongqing University of Technology：Natural Science

基　　金：国家自然科学基金项目(11871124,11471060);重庆市基础研究与前沿探索项目(cstc2018jcyjAX0241);重庆理工大学研究生创新项目(ycx20192082)。

摘　　要：考虑了分层设计下的无金标准部分核实数据对疾病流行率的小样本齐性检验,基于2种模型提出了加权最小二乘统计量、基于对数变换、logit变换、双对数变换的加权最小二乘统计量和Score统计量,并基于这些统计量提出了基于Bootstrap的检验过程。模拟研究比较了基于大样本的渐近检验过程和基于Bootstrap的小样本检验过程。研究结果表明:在小样本下,基于Bootstrap方法的检验使其犯第一类错误的概率更接近名义水平且具有更高的功效,基于Score统计量的Bootstrap检验过程在2种模型下都表现良好,因而被实际应用所推荐。实例分析进一步验证了所提方法的有效性。In this paper,the homogeneity testing of disease prevalence rates for stratified partially validated data without gold-standard under sample sizes is considered.Under two models,the weightleast-squares test,the weight-least-squares tests based on log,logit and double-log transformation,and Score test are proposed.The bootstrap test procedures based on these test statistics are developed for the homogeneity testing under sample sizes.The comparison of asymptotic method and the Bootstrap method is evaluated by simulation study for sample size designs.The simulation results show that Bootstrap procedures perform better than the corresponding asymptotic procedures in the sense that the former empirical type I error is closer to the nominal level than the latter,and the Score test based on Bootstrap method performs well under two models,and hence be recommended to practical applications.A real data is used to illustrate the proposed methods.

关 键 词：BOOTSTRAP方法 齐性检验 部分核实数据 分层设计 SCORE统计量 

分 类 号：R181[医药卫生—流行病学] O212.1[医药卫生—公共卫生与预防医学]