Interval-censored survival time data: confidence intervals for the non-parametric survivor function.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Inference

Background:

• Survival data analysis is crucial in many fields.
• Interval-censored data, where failure times are known only within intervals, presents unique analytical challenges.
• Accurate confidence intervals are essential for reliable interpretation of survival function estimates.

Purpose of the Study:

• To evaluate three distinct methods for calculating pointwise confidence intervals.
• To assess the performance of these methods when applied to non-parametric survivor function estimation from interval-censored data.
• To identify the most accurate and reliable method for confidence interval construction in this context.

Main Methods:

• The study assessed three methods: full information matrix, modified information matrix (with zero estimates deleted), and likelihood ratio inference.
• A simulation study was conducted to compare the performance of these methods.
• Pointwise confidence intervals for the non-parametric survivor function were calculated.

Main Results:

• The likelihood ratio inference method demonstrated superior accuracy in confidence interval calculations.
• Confidence intervals generated by the likelihood ratio method consistently maintained coverage near the nominal 95% level.
• The other two methods showed less consistent or accurate coverage compared to the likelihood ratio approach.

Conclusions:

• The likelihood ratio method is recommended for constructing confidence intervals for non-parametric survivor functions with interval-censored data.
• This method offers reliable and accurate estimation, ensuring statistical validity.
• Accurate confidence intervals are vital for robust survival data interpretation.