topic Re: Proc Mianalyze Computing Overall p-value for Supremum Test in Proportional Hazards Assumption in Statistical Procedures

Proc Mianalyze Computing Overall p-value for Supremum Test in Proportional Hazards Assumption

YYK273 — Tue, 07 May 2024 18:26:31 GMT

I'm currently working on testing the overall Proportional Hazards Assumption for my 5 multiple imputation datasets.

My challenge lies in determining the overall p-value of the Supremum Test for Proportional Hazards Assumption. While the test only provides the Maximum Absolute Value and Pr > MaxAbsVal, I'm uncertain about how to derive the overall p-value from proc mianalyze.

Could you kindly provide some guidance or insights on how to approach this issue? Any assistance or pointers you could offer would be greatly appreciated.

Thank you very much for your time and assistance.

Re: Proc Mianalyze Computing Overall p-value for Supremum Test in Proportional Hazards Assumption

SAS_Rob — Wed, 08 May 2024 14:26:08 GMT

The theory behind combining the multiply imputed estimates requires that there be a point estimate and standard error. Since the Supremum test only reports the maximum and a p-value, there is no easy way to get combined estimates.

Re: Proc Mianalyze Computing Overall p-value for Supremum Test in Proportional Hazards Assumption

Season — Sun, 07 Jul 2024 10:39:16 GMT

You are another one to have discovered a limitation of the MIANALYZE procedure and a field on which little research has been conducted-i.e., pooling P values in multiple imputation.

You can refer to my reply to another thread, which also concerns pooling P values in multiple imputation: Re: Adjusted pvalue with Proc Mianalyze - SAS Support Communities.