We propose Bayesian nonparametric Weibull delegate racing (WDR) for survival analysis with competing events and achieve both model interpretability and flexibility. Utilizing a natural mechanism of surviving competing events, we assume a race among a potentially infinite number of sub-events. In doing this, WDR accommodates nonlinear covariate effects with no need of data transformation. Moreover, WDR is able to handle left truncation, time-varying covariates, different types of censoring, and missing event times or types. We develop an efficient MCMC algorithm based on Gibbs sampling for Bayesian inference and provide an \texttt{R} package. Synthetic data analysis and comparison with benchmark approaches demonstrate WDR's outstanding performance and parsimonious nonlinear modeling capacity. In addition, we analyze two real data sets and showcase advantages of WDR. Specifically, we study time to death of three types of lymphoma and show the potential of WDR in modeling nonlinear covariate effects and discovering new diseases. We also use WDR to investigate the age at onset of mild cognitive impairment and interpret the accelerating or decelerating effects of biomarkers on the progression of Alzheimer's disease.

翻译：我们提出贝叶斯非参数韦布尔委托竞赛（WDR）方法，用于含竞争事件的生存分析，同时实现模型的可解释性与灵活性。利用生存竞争事件的自然机制，我们假设存在潜在无限多个子事件之间的竞争过程。通过这一设计，WDR无需数据变换即可处理非线性协变量效应。此外，该方法能够应对左截断、时变协变量、多种删失类型以及缺失事件时间或类型等问题。我们开发了基于吉布斯采样的高效MCMC算法用于贝叶斯推断，并提供R语言软件包。合成数据分析及与基准方法的对比表明，WDR具有卓越的性能与简约的非线性建模能力。同时，我们分析了两组真实数据以展示WDR的优势：其一，针对三种淋巴瘤的死亡时间研究，揭示了WDR在非线性协变量效应建模及新疾病发现方面的潜力；其二，利用WDR探究轻度认知障碍的发病年龄，阐释了生物标志物对阿尔茨海默病进展的加速或减缓效应。