This study focuses on estimating a nonparametric regression model with right-censored data when the covariate is subject to measurement error. To achieve this goal, it is necessary to solve the problems of censorship and measurement error ignored by many researchers. Note that the presence of measurement errors causes biased and inconsistent parameter estimates. Moreover, non-parametric regression techniques cannot be applied directly to right-censored observations. In this context, we consider an updated response variable using the Buckley–James method (BJM), which is essentially based on the Kaplan–Meier estimator, to solve the censorship problem. Then the measurement error problem is handled using the kernel deconvolution method, which is a specialized tool to solve this problem. Accordingly, three denconvoluted estimators based on BJM are introduced using kernel smoothing, local polynomial smoothing, and B-spline techniques that incorporate both the updated response variable and kernel deconvolution.The performances of these estimators are compared in a detailed simulation study. In addition, a real-world data example is presented using the Covid-19 dataset.

本研究的重点是当协变量存在测量误差时，使用右删失数据估计非参数回归模型。为了实现这一目标，需要解决被许多研究者忽视的审查和测量误差问题。请注意，测量误差的存在会导致参数估计出现偏差和不一致。此外，非参数回归技术不能直接应用于右删失观测。在这种情况下，我们考虑使用巴克利-詹姆斯方法（BJM）更新响应变量，该方法本质上基于卡普兰-迈耶估计器，以解决审查问题。然后使用核反卷积方法处理测量误差问题，这是解决该问题的专用工具。因此，使用核平滑、局部多项式平滑和 B 样条技术引入了基于 BJM 的三种反卷积估计器，其中结合了更新的响应变量和核反卷积。在详细的模拟研究中对这些估计器的性能进行了比较。此外，还使用 ​​Covid-19 数据集提供了一个真实世界的数据示例。