Jiang Yue, Lieber Sarah R
Department of Statistical Science, Duke University, Durham, North Carolina, USA.
Department of Medicine, Division of Digestive and Liver Diseases, University of Texas Southwestern (UTSW) Medical Center, Dallas, Texas, USA.
Liver Transpl. 2025 Feb 1;31(2):221-230. doi: 10.1097/LVT.0000000000000451. Epub 2024 Aug 6.
We present a tutorial on quantile regression, an underutilized yet valuable class of multivariable linear regression models that allows researchers to understand more fully the conditional distribution of response as compared to models based on conditional means. Quantile regression models are flexible, have attractive interpretations, and are implemented in most statistical software packages. Our focus is on an intuitive understanding of quantile regression models, particularly as compared with more familiar regression methods such as conditional mean models as estimated using ordinary least squares (OLS). We frame our tutorial through 2 clinical case studies in the field of liver transplantation: one in the context of estimating the recipient's financial burden after transplantation and another in estimating intraoperative blood transfusion needs. Our real-world cases demonstrate how quantile regression models give researchers a richer understanding of relationships in the data and provide a more nuanced clinical understanding compared to more commonly used linear regression models. We encourage researchers to explore quantile regression as a tool to answer relevant clinical research questions and support their more widespread adoption.
我们提供了一篇关于分位数回归的教程，分位数回归是一类未得到充分利用但很有价值的多变量线性回归模型，与基于条件均值的模型相比，它能让研究人员更全面地了解响应的条件分布。分位数回归模型灵活多变，具有吸引人的解释力，并且在大多数统计软件包中都能实现。我们的重点是直观地理解分位数回归模型，特别是与更熟悉的回归方法（如使用普通最小二乘法（OLS）估计的条件均值模型）进行比较。我们通过肝移植领域的两个临床案例研究来构建我们的教程：一个是估计移植后受者的经济负担，另一个是估计术中输血需求。我们的实际案例展示了分位数回归模型如何使研究人员更深入地理解数据中的关系，并且与更常用的线性回归模型相比，能提供更细致入微的临床理解。我们鼓励研究人员探索将分位数回归作为一种工具来回答相关的临床研究问题，并支持其更广泛地应用。
