Promoting sign consistency in the cure model estimation and selection.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Two-part cure rate models are used when some subjects have very long survival times.
• These models estimate cure probability (binary response) and survival time (survival response) separately.
• Existing methods often ignore interconnections between cure and survival, or use restrictive proportionality constraints.

Purpose of the Study:

• To propose a new method for estimating two-part cure rate models.
• To introduce a sign-based penalty for qualitative similarity between model components.
• To improve variable selection and handle high-dimensional data in cure rate modeling.

Main Methods:

• Developed a two-part cure rate model with a sign-based penalty.
• Applied the penalty to promote qualitative similarity between cure and survival probabilities.
• Incorporated variable selection and high-dimensional data accommodation via penalties.

Main Results:

• The proposed sign-based penalty offers a less restrictive approach than existing proportionality methods.
• The strategy is more informative by considering the interconnections between cure and survival.
• Numerical studies, including simulations and data analyses, confirmed the proposed method's superior performance.

Conclusions:

• The novel sign-based penalty effectively improves two-part cure rate model estimation.
• This approach provides a more flexible and informative alternative for survival analysis with long-term survivors.
• The method demonstrates advantages in variable selection and high-dimensional settings.