报告题目：Conditional Independence Test via Classifier
时 间：2026年11月13日（星期五）9：00
地 点：理工南楼201
主 办：数学与统计学院、分析数学及应用教育部重点实验室、统计学与人工智能福建省高校重点实验室
参加对象：感兴趣的老师和学生
报告摘要：Testing conditional independence of continuous variables is a fundamental yet challenging task in statistics, particularly in high-dimensional settings where distributional or structural assumptions are unavailable. In this work, we reformulate the conditional independence test as a problem of testing whether two samples are drawn from the same distribution. To circumvent the practical hurdle that one sample set is often infeasible to generate directly, we introduce a k-nearest-neighbor local sampling strategy. Moreover, to address the failure of rank-based tests in high dimensions, we replace the conventional rank sum test with a classifier-based approach. This allows us to leverage powerful machine learning classification algorithms, thereby enabling application to complex, high dimensional data. Moreover, our test statistic is asymptotically normal under the null, given certain regularity conditions. We demonstrate the performance of our method on both simulated and real datasets.
报告人简介：陈夏，陕西师范大学数学与统计学院教授、博士生导师。武汉大学概率论与数理统计博士，北京师范大学统计学博士后。兼任陕西省统计学学会副理事长、陕西省工业与应用数学学会副理事长和中国现场统计研究会多个分会的常务理事。主要从事高维数据统计分析和概率极限理论方面的研究。在国内外统计学重要学术期刊发表论文40余篇，在科学出版社出版专著1部、教材1部。主持国家级和省部级项目多项。获陕西省高等教育优秀教材一等奖、陕西省学位与研究生教育成果奖一等奖和陕西高校科学技术奖二等奖等奖励多项。