An adjustment to improve the bivariate survivor function repaired NPMLE.

Area of Science:

• Statistics
• Survival Analysis

Background:

• Bivariate survivor function estimation is crucial for analyzing time-to-event data.
• Existing methods like the repaired nonparametric maximum likelihood estimator (NPMLE) have limitations in practical application, including sensitivity to bandwidth and poor variance estimation agreement.

Purpose of the Study:

• To present a modified repaired NPMLE that enhances performance in moderate sample sizes.
• To reduce the sensitivity of the estimator to bandwidth selection.
• To provide more extensive simulation results and apply the methods to real-world data.

Main Methods:

• Proposed a modified repaired NPMLE based on a hazard function mapping for truncated failure time variates.
• Conducted extensive simulation studies to compare the modified estimator with existing methods.
• Evaluated Greenwood-like variance estimates for the modified estimator.

Main Results:

• The modified repaired NPMLE demonstrates improved performance in moderate sample sizes compared to the original NPMLE.
• The modified estimator exhibits reduced sensitivity to bandwidth selection.
• Simulation studies show better agreement for variance estimates with the modified approach.

Conclusions:

• The modified repaired NPMLE offers a more robust and practical approach for bivariate survival function estimation.
• This enhancement addresses key limitations of the original NPMLE, making it more suitable for real-world applications in survival analysis.