Luo Jingqin, Gao Feng, Liu Jingxia, Wang Guoqiao, Chen Ling, Fagan Anne M, Day Gregory S, Vöglein Jonathan, Chhatwal Jasmeer P, Xiong Chengjie
Siteman Cancer Center Biostatistics Shared Resource, Division of Public Health Sciences, Department of Surgery, Washington University School of Medicine, St. Louis, MO, USA.
Division of Biostatistics, Washington University School of Medicine, St. Louis, MO, USA.
J Appl Stat. 2021 Mar 18;49(9):2246-2270. doi: 10.1080/02664763.2021.1899141. eCollection 2022.
Bivariate correlation coefficients (BCCs) are often calculated to gauge the relationship between two variables in medical research. In a family-type clustered design where multiple participants from same units/families are enrolled, BCCs can be defined and estimated at various hierarchical levels (subject level, family level and marginal BCC). Heterogeneity usually exists between subject groups and, as a result, subject level BCCs may differ between subject groups. In the framework of bivariate linear mixed effects modeling, we define and estimate BCCs at various hierarchical levels in a family-type clustered design, accommodating subject group heterogeneity. Simplified and modified asymptotic confidence intervals are constructed to the BCC differences and Wald type tests are conducted. A real-world family-type clustered study of Alzheimer disease (AD) is analyzed to estimate and compare BCCs among well-established AD biomarkers between mutation carriers and non-carriers in autosomal dominant AD asymptomatic individuals. Extensive simulation studies are conducted across a wide range of scenarios to evaluate the performance of the proposed estimators and the type-I error rate and power of the proposed statistical tests. Abbreviations: BCC: bivariate correlation coefficient; BLM: bivariate linear mixed effects model; CI: confidence interval; AD: Alzheimer's disease; DIAN: The Dominantly Inherited Alzheimer Network; SA: simple asymptotic; MA: modified asymptotic.
双变量相关系数（BCCs）常用于医学研究中衡量两个变量之间的关系。在家族型聚类设计中，当同一单位/家族的多个参与者被纳入时，可以在不同层次水平（个体水平、家族水平和边际BCC）定义和估计BCCs。个体组之间通常存在异质性，因此，个体水平的BCCs在个体组之间可能会有所不同。在双变量线性混合效应建模框架下，我们在家族型聚类设计的不同层次水平定义和估计BCCs，以适应个体组异质性。构建了简化和修正的渐近置信区间用于BCC差异，并进行了 Wald 型检验。对一项现实世界中的阿尔茨海默病（AD）家族型聚类研究进行分析，以估计和比较常染色体显性AD无症状个体中突变携带者和非携带者之间已确立的AD生物标志物的BCCs。在广泛的场景中进行了大量模拟研究，以评估所提出估计量的性能以及所提出统计检验的I型错误率和检验效能。缩写：BCC：双变量相关系数；BLM：双变量线性混合效应模型；CI：置信区间；AD：阿尔茨海默病；DIAN：显性遗传阿尔茨海默病网络；SA：简单渐近；MA：修正渐近。
