Nonparametric modeling of the gap time in recurrent event data

Area of Science:

• Biostatistics
• Survival Analysis
• Biomedical Data Analysis

Background:

• Recurrent event data are common in biomedical and engineering fields, where events can happen multiple times.
• Accurate estimation of the hazard function for gap times is crucial for understanding these repeated events.

Purpose of the Study:

• To develop a nonparametric method for estimating the hazard function for gap times in recurrent event data.
• To incorporate covariates into the hazard function estimation.
• To provide methods for smoothing parameter selection and confidence interval derivation.

Main Methods:

• A penalized likelihood model is proposed for hazard function estimation.
• Subject-wise cross-validation is used for smoothing parameter selection.
• Bayes model is employed for deriving confidence intervals.
• Eigenvalue analysis is used to establish asymptotic convergence rates.

Main Results:

• The penalized likelihood model provides an effective way to estimate hazard functions for gap times with covariates.
• The proposed smoothing parameter selection method is reliable.
• Asymptotic convergence rates of estimates are theoretically established.
• Empirical studies demonstrate the method's performance.

Conclusions:

• The developed penalized likelihood approach offers a robust technique for analyzing recurrent event data.
• The method is validated through application to bladder tumor cancer data, showing its practical utility.