What question did this study set out to answer?

Compare the validity and efficiency of three common approximation methods (Breslow, Kalbfleisch-Prentice, and Efron) for handling tied event times in Cox proportional hazards models.

September 1, 1997Biometrics

Validity and Efficiency of Approximation Methods for Tied Survival Times in Cox Regression

View Full Paper

Ask AI

Bookmark

Authors

IHIrva Hertz‐PicciottoSlovak University of Technology in Bratislava BRBeverly RockhillUniversity of North Carolina at Greensboro

Discussion

Member takes

Implication

Simulation study demonstrates superior accuracy and lower bias with Efron's tie approximation across sample sizes, suggesting it should replace Breslow as standard default.

Key Points

Compare the validity and efficiency of three common approximation methods (Breslow, Kalbfleisch-Prentice, and Efron) for handling tied event times in Cox proportional hazards models.
Simulated a two-sample survival problem with exponential hazard, constant hazard ratio, no censoring, and sample sizes of n = 25, 50, 250, and 500 per group.
Grouped failure times into light (mean 2.5), moderate (mean 5.0), and heavy (mean 10.0) ties per interval.
Evaluated model performance and confidence interval coverage across all sample sizes and tie frequencies using Breslow, Kalbfleisch-Prentice, and Efron approximations.
The Breslow method systematically underestimated beta while Kalbfleisch-Prentice overestimated beta, with bias worsening as tie density increased.
The Efron approximation performed best across all conditions, maintaining bias under 2% for n = 25 per group and under 1% for sample sizes larger than 50 per group.
Confidence interval coverage ranged from 94% to 96% for all estimators, but tail probabilities were asymmetric for Breslow and Kalbfleisch-Prentice while Efron remained closer to the nominal 2.5% tail probability.

Cite This Study

Hertz‐Picciotto et al. (1997) studied this question.

synapsesocial.com/papers/6a15604d5347fbb1739fae23 https://doi.org/10.2307/2533573

View Full Paper

Ask AI

Bookmark

Implication

Simulation study demonstrates superior accuracy and lower bias with Efron's tie approximation across sample sizes, suggesting it should replace Breslow as standard default.

Validity and Efficiency of Approximation Methods for Tied Survival Times in Cox Regression

Discussion

Member takes

Implication

Key Points

Cite This Study

Validity and Efficiency of Approximation Methods for Tied Survival Times in Cox Regression

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider