Testing the proportional hazards regression model against some general alternatives

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• The proportional hazards regression model is a cornerstone of survival data analysis.
• Assessing the goodness-of-fit is crucial for validating model assumptions.
• Existing methods may not adequately capture time-varying covariate effects.

Purpose of the Study:

• To propose a novel class of goodness-of-fit tests for the proportional hazards regression model.
• To develop tests sensitive to time-dependent parameter variations.
• To provide a framework for assessing model fit under specific alternative hypotheses.

Main Methods:

• Development of goodness-of-fit statistics based on deviations from the proportional hazards assumption.
• Consideration of alternatives where model parameters change as step functions of time.
• Formulation of specific tests by restricting the form of time-dependent parameter changes.
• Outline of computational procedures for implementing the proposed tests.

Main Results:

• A class of goodness-of-fit tests is formally proposed.
• The tests are shown to be sensitive to time-varying parameters modeled as step functions.
• Restricted forms of step functions lead to more powerful tests against specific alternatives.
• Computational details for practical application are provided.

Conclusions:

• The proposed tests offer a valuable tool for assessing the proportional hazards assumption in regression models.
• These methods enhance the reliability of survival analyses by detecting time-dependent effects.
• The framework allows for tailored goodness-of-fit assessments based on expected deviations from the proportional hazards assumption.