A specification test for univariate and multivariate proportional hazards models

Area of Science:

• Statistics
• Biostatistics
• Econometrics

Background:

• Model specification testing is crucial for reliable statistical inference.
• Censored survival data analysis commonly employs proportional hazards models.
• Existing methods for model specification in survival analysis have limitations.

Purpose of the Study:

• To apply White's information matrix (IM) test for model specification to proportional hazards models.
• To examine alternative estimators of the IM test statistic and their size performance.
• To illustrate the utility of the IM test for both univariate and multivariate censored survival data.

Main Methods:

• Application of White's information matrix (IM) test.
• Evaluation of alternative estimators for the test statistic.
• Examination of the size performance of these estimators.
• Integration of the IM test within the independence working model (IWM) for multivariate data.

Main Results:

• The IM test is demonstrated to be applicable to proportional hazards models.
• Alternative estimators of the IM test statistic show varying size performance.
• The IM test can be effectively used within the IWM framework for multivariate survival data.

Conclusions:

• White's IM test provides a valuable tool for assessing the correct specification of proportional hazards models.
• The study offers practical guidance on applying the IM test to univariate and multivariate censored survival data.
• The findings support the use of the IM test as part of robust statistical modeling strategies in survival analysis.