Scholarlink

<p>Establishing whether a variable <i>Y</i> is a function of a variable <i>X</i> is a fundamental problem in statistical analysis. Recently, a novel coefficient known as Chatterjee’s correlation has been introduced to address this problem of testing for functional dependencies. This correlation is now widely used to assess dependencies between variables. However, statistical studies frequently encounter missing data. Failing to account for these missing values can lead to biased and misleading results. To address this issue, we consider the scenario where the variable <i>Y</i> contains data that are missing at random. We then extend Chatterjee’s correlation by incorporating inverse probability weighting. A theoretical framework is established to study the asymptotic properties of the proposed estimator. Our method demonstrates strong performance in finite-sample simulations and remains robust to the functional form of the dependency. Simulation studies confirm the excellent performance of the proposed estimator with finite sample sizes, even under varying rates of missing data. Furthermore, we demonstrate its practical utility by applying it to real-world datasets where the variable <i>Y</i> is subject to missingness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Improvement of Chatterjee’s correlation with missing at random data in the Y variable

  • 期刊论文

  • 发表日期：

    2026年4月21日

查看全文
  • Fayyaz Bahari

摘要

Establishing whether a variable Y is a function of a variable X is a fundamental problem in statistical analysis. Recently, a novel coefficient known as Chatterjee’s correlation has been introduced to address this problem of testing for functional dependencies. This correlation is now widely used to assess dependencies between variables. However, statistical studies frequently encounter missing data. Failing to account for these missing values can lead to biased and misleading results. To address this issue, we consider the scenario where the variable Y contains data that are missing at random. We then extend Chatterjee’s correlation by incorporating inverse probability weighting. A theoretical framework is established to study the asymptotic properties of the proposed estimator. Our method demonstrates strong performance in finite-sample simulations and remains robust to the functional form of the dependency. Simulation studies confirm the excellent performance of the proposed estimator with finite sample sizes, even under varying rates of missing data. Furthermore, we demonstrate its practical utility by applying it to real-world datasets where the variable Y is subject to missingness.