Related Experiment Videos
Comparing proportional hazards and accelerated failure time models for survival analysis
Jesus Orbe1, Eva Ferreira, Vicente Núñez-Antón
1Departamento de Econometría y Estadística Facultad de Ciencias Económicas y Empresariales, Universidad del País Vasco Euskal/Herriko Unibertsitatea, Bilbao, Spain.
Statistics in Medicine
|October 31, 2002
Summary
This study introduces a novel censored linear regression method for survival analysis. It offers accurate estimation without distribution assumptions or proportional hazards, outperforming traditional models in simulations and real-world data.
Area of Science:
- Statistics
- Biostatistics
- Survival Analysis
Background:
- Censored data is common in survival analysis.
- Cox proportional hazards models are widely used but rely on specific assumptions.
- Existing methods may require distributional assumptions or proportional hazards, limiting their applicability.
Purpose of the Study:
- To propose a new method for censored linear regression in survival analysis.
- To offer an alternative to Cox models when the proportional hazards assumption is violated.
- To provide a method that does not require knowledge of the duration variable's distribution.
Main Methods:
- Development of a censored linear regression model.
- Estimation and inference without assuming the distribution of the duration variable.
- Application to two real-world datasets and a simulation study for performance analysis.
Main Results:
- The proposed method allows estimation and inference without distributional assumptions.
- It does not require the proportional hazards assumption, offering flexibility.
- Performance analysis through real examples and simulations indicates more precise results compared to Cox and AFT models.
Conclusions:
- The new censored linear regression method is a viable and potentially more precise alternative for survival analysis.
- Its simplicity in implementation and interpretation enhances its practical utility.
- It addresses limitations of traditional models, particularly when assumptions are not met.