The sparse estimation of the semiparametric linear transformation model with dependent current status data

Area of Science:

• Biostatistics
• Statistical modeling
• Survival analysis

Background:

• Current status data (Type I interval-censored data) presents unique challenges in statistical analysis.
• Failure times in such data can be dependent on censoring times, with unspecified association.
• Existing methods may not adequately address sparse estimation under these complex dependencies.

Purpose of the Study:

• To develop a robust statistical method for sparse estimation in semiparametric linear transformation models.
• To accurately estimate the association parameter between failure and censoring times.
• To apply the developed method to a real-world Alzheimer's disease study.

Main Methods:

• Utilized copula models to capture the dependence between failure and censoring times.
• Employed a two-stage estimation procedure for association and regression parameters.
• Implemented penalized maximum likelihood estimation with broken adaptive ridge regression.
• Applied Bernstein polynomials for approximating nonparametric functions.

Main Results:

• Established the oracle property of the proposed estimation method, ensuring asymptotic efficiency.
• Numerical simulations demonstrated the method's effectiveness in practical scenarios.
• Successfully applied the methodology to analyze data from an Alzheimer's disease study.

Conclusions:

• The proposed method provides a powerful tool for sparse estimation with interval-censored data.
• The approach effectively handles unspecified dependence structures between failure and censoring times.
• This work offers valuable insights and tools for analyzing complex health-related data, such as in Alzheimer's research.