Computing P-values for a class of permutation tests of equal survival functions

Area of Science:

• Biostatistics
• Survival Analysis
• Medical Statistics

Background:

• Permutation tests are valuable for comparing survival distributions with censored data.
• Existing methods may be computationally intensive, limiting practical application.
• Medical studies often involve paired and unpaired censored survival time data.

Purpose of the Study:

• To provide practical computing routines for permutation tests proposed by Dallas and Rao (2000).
• To facilitate the implementation of powerful tests for comparing two survival distributions.
• To address the computational challenges associated with increasing sample sizes in survival data analysis.

Main Methods:

• Development and provision of computing routines for specific permutation tests.
• Focus on tests designed for randomly right censored survival time data.
• Inclusion of methods that handle both paired and unpaired observations.

Main Results:

• The provided routines enable efficient execution of the advocated permutation tests.
• These tests demonstrate high power for detecting scale and location shifts.
• The methods show improved power compared to paired tests that exclude unpaired data.

Conclusions:

• The computing routines make advanced permutation tests for censored survival data accessible.
• These tests are particularly useful in medical settings with mixed data types.
• Practical implementation is now feasible, allowing for more robust survival distribution comparisons.