Computation of nonparametric convex hazard estimators via profile methods

Area of Science:

• Statistics
• Survival Analysis
• Computational Statistics

Background:

• Estimating hazard functions is crucial in survival analysis.
• Nonparametric methods offer flexibility but can be computationally intensive.
• Convexity constraints can improve estimation accuracy.

Purpose of the Study:

• To develop an efficient algorithm for computing the nonparametric maximum likelihood estimator (NPMLE) of a convex hazard function.
• To address the computational challenges in hazard function estimation.

Main Methods:

• A two-step maximization approach is proposed.
• The support reduction algorithm is used for initial maximization.
• A bisection algorithm is applied to the profile likelihood for global maximization.

Main Results:

• The profile likelihood is shown to be quasi-concave with respect to the antimode.
• The proposed algorithm efficiently computes the NPMLE for convex hazard functions.
• The algorithm's performance is demonstrated on artificial and real-world datasets.

Conclusions:

• The novel profile likelihood algorithm provides an efficient method for convex hazard function estimation.
• The approach is validated using diverse datasets, including Canadian male and female lifetime data.
• This method enhances the accuracy and applicability of survival data analysis.