Langholz B, Thomas D C
Department of Preventive Medicine, University of Southern California, School of Medicine 90033-9987.
Biometrics. 1991 Dec;47(4):1563-71.
Cohort sampling designs are proposed which one would intuitively expect to be more efficient than nested case-control sampling. Two of these designs start with a nested case-control sample and distribute controls to sampled risk sets other than those for which they were picked. The third design has the goal of maximizing the number of distinct persons in a nested case-control sample. Simulation results show surprisingly little gain, and more often a loss in efficiency of these new designs relative to nested case-control sampling. This is due to the sampling-induced covariance between score terms. We conclude that the often stated intuition that nested case-control sampling does not make good use of sampled individuals' covariate histories is false.
提出了队列抽样设计，直观地看，人们会认为这些设计比巢式病例对照抽样更有效。其中两种设计从巢式病例对照样本开始，将对照分配到除了最初选取对照时所在风险集之外的其他抽样风险集中。第三种设计的目标是使巢式病例对照样本中不同个体的数量最大化。模拟结果显示，相对于巢式病例对照抽样，这些新设计的效率提升惊人地小，而且更多时候是效率降低。这是由于得分项之间抽样引起的协方差。我们得出结论，那种认为巢式病例对照抽样没有充分利用抽样个体协变量历史的常见直觉是错误的。
