Inferring Bivariate Association from Respondent-driven Sampling Data
针对受访者驱动抽样中个体间抽样依赖导致标准统计检验失效的问题，提出一种半参数方法估计零分布，实现更有效的双变量关联检验，并应用于纽约市年轻非法阿片类药物使用者特征研究。
Abstract Respondent-driven sampling (RDS) is an effective method of collecting data from many hard-to-reach populations. Valid statistical inference for these data relies on many strong assumptions. In standard samples, we assume observations from pairs of individuals are independent. In RDS, this assumption is violated by the sampling dependence between individuals. We propose a method to semi-parametrically estimate the null distributions of standard test statistics in the presence of sampling dependence, allowing for more valid statistical testing for dependence between pairs of variables within the sample. We apply our method to study characteristics of young adult illicit opioid users in New York City.