Change point detection in Cox proportional hazards mixture cure model

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Mixture cure models address survival data where some subjects never experience the event.
• Standard models may not capture abrupt changes in covariate effects over time.

Purpose of the Study:

• To develop a Cox proportional hazards mixture cure model incorporating change point analysis.
• To estimate covariate effects that may change abruptly at a specific threshold.

Main Methods:

• Utilized nonparametric maximum likelihood estimation for semiparametric estimates.
• Employed a two-step Expectation-Maximization algorithm for estimation.
• Applied m out of n bootstrap and Louis algorithm for standard error estimation.

Main Results:

• Established consistency, convergence rates, and asymptotic distributions of estimators.
• Demonstrated n-consistency for the change point estimator.
• Developed a test for change point existence.

Conclusions:

• The proposed method effectively models survival data with change points in covariate effects.
• The methodology provides robust estimation and hypothesis testing for mixture cure models.
• Validated through simulations and real-world data analysis.