针对 COVID-19 抽样偏差的一种简单校正
统计方法学
2022-02-24 v3 种群与进化
摘要
COVID-19 检测已成为估计患病率的常用方法，进而辅助公共卫生决策以遏制与减缓疾病传播。所采用的抽样设计常有偏差，因其未反映真实的潜在人群。例如，有强烈症状者比无症状者更可能被检测。这导致患病率的偏差估计（过高）。典型的抽样后校正并非总可行。此处我们提出一种简单的偏差校正方法，其推导并改编自荟萃分析中针对发表偏差的校正。该方法具足够通用性以允许多种定制，从而在实践中更为有用。利用已收集信息即可轻松实现。通过一次模拟与两个真实数据集，我们显示偏差校正可大幅降低估计误差。
引用
@article{arxiv.2007.07426,
title = {A simple correction for COVID-19 sampling bias},
author = {Daniel Andrés Díaz-Pachón and J Sunil Rao},
journal= {arXiv preprint arXiv:2007.07426},
year = {2022}
}
备注
14 pages. Title changed. The whole Section 7 with information from Lombardy, Italy, was added (another real dataset). Some typos were corrected. In spite of several lengthy additions, no substantial changes were done to the paper. The goal of the additions was more to clarify than to correct