用合成统计量来整合推断结果

Abstract: Learning from the collective wisdom of crowds parallels the statistical concept of fusion learning from multiple data sources or studies. However, integrating inferences from diverse sources poses significant challenges due to cross-source heterogeneity and data-sharing limitations. Studies often rely on varied designs and modeling techniques, and stringent data privacy norms can prohibit even the sharing of summary statistics. In this talk, we will discuss the construction of "synthetic statistics" that mimic the summary statistics used for inference, enabling the fusion of inference results from multiple sources.

CV: 刚博文，复旦大学管理学院助理教授，2020年博士毕业于南加州大学。他的研究方向包括多重假设检验，序贯分析，无模型推理方法等。在JASA，JMLR, Biometrtika等期刊上发表多篇论文。