Reliable survival analysis based on the Dirichlet process

Area of Science:

• Statistics
• Biostatistics
• Survival Analysis

Background:

• Estimating survival functions from right-censored data is crucial in many fields.
• Traditional Dirichlet process priors often require strong assumptions or complex parameter elicitation.
• Robustness in survival analysis is essential for reliable inference.

Purpose of the Study:

• To develop a robust Dirichlet process for survival function estimation with right-censored data.
• To provide a near-ignorance prior approach that minimizes assumptions about lifetime distributions.
• To enable robust inferences and sensitivity analyses for censored lifetime data.

Main Methods:

• Utilizing a robust Dirichlet process with a near-ignorance prior.
• Developing a nonparametric estimator for survival probability.
• Formulating a hypothesis test for comparing lifetimes between two populations.
• Implementing methods for sensitivity analysis regarding prior-dependent decisions.

Main Results:

• A robust method for estimating survival functions from right-censored data was established.
• A nonparametric survival probability estimator was derived.
• A hypothesis test for comparing population lifetimes was developed.
• The approach demonstrated robustness on simulated and real-world (Australian AIDS survival) datasets.

Conclusions:

• The robust Dirichlet process offers a flexible and assumption-light framework for survival analysis.
• The developed methods provide reliable tools for survival probability estimation and hypothesis testing.
• The R package facilitates the application of these robust statistical techniques.