Wang Jianing, Kline David M, White Laura Forsberg
Biostatistics Center, Massachusetts General Hospital, Boston, MA, USA.
Department of Medicine, Harvard Medical School, Boston, MA, USA.
Stat Methods Med Res. 2024 Oct;33(10):1818-1835. doi: 10.1177/09622802241275413. Epub 2024 Sep 30.
Approaches to population size estimation are of importance across a wide spectrum of disciplines, especially when census and simple random sampling are impractical. The capture-recapture method and the multiplier-benchmark method are two commonly used approaches that use data that partially capture the target population and overlap in a known way. Due to similarities in required data structures, the approaches are often used interchangeably without a critical appraisal of the underlying assumptions, especially in the two-sample case. Here, we describe the similarities and differences of the sampling mechanisms and assumptions underlying both approaches. We emphasize that the capture-recapture method assumes data sources as random samples and describes two-way inclusion histories, while in multiplier-benchmark method, one source captures a fixed sub-population, and the one-way inclusion histories are modeled. We also discuss the implications of these differences through simulation and real data to guide the choice of method in practice. A careful study of the data structures, relationships, and data generation processes is crucial for assessing the appropriateness of using these methods.
种群规模估计方法在广泛的学科领域中都很重要，尤其是在进行普查和简单随机抽样不切实际的情况下。捕获-再捕获方法和乘数-基准方法是两种常用的方法，它们使用的数据部分捕获目标种群，并以已知方式重叠。由于所需数据结构的相似性，这些方法经常互换使用，而没有对基本假设进行批判性评估，尤其是在双样本情况下。在此，我们描述了这两种方法背后的抽样机制和假设的异同。我们强调，捕获-再捕获方法将数据源假定为随机样本，并描述双向包含历史，而在乘数-基准方法中，一个数据源捕获固定的子种群，并对单向包含历史进行建模。我们还通过模拟和实际数据讨论了这些差异的影响，以指导实践中方法的选择。仔细研究数据结构、关系和数据生成过程对于评估使用这些方法的适当性至关重要。
