Cox regression: "stphplot" and "estat phtest, detail" difference

[Soyun Park]

Hello,

I've been working on a cox regression analysis and faced an issue in assessing the ph assumption.
I first tested my assumption using "estat phtest, detail" and my core independent variable got a Prob>chi2 of 0.4021, thus the PH assumption holds.
However, when I used the "stphplot" to see how that independent variable looks graphically the lines cross at a point, meaning they are not parallel, which goes against what the proportional hazard cox regression requires.

Which do I follow on with?
If the global test still holds for the estat phtest significance test approach is it still okay to proceed?
Also, if I have to follow on with the graphical examination and there is an issue with my independent variable, is there a survival analysis I can use with non-proportional hazard?

Thank you so much in advance!!

[Marcos Almeida]

Welcome to the Stata Forum / Statalist,


In case the model violates the PH assumption, you may need to apply some strategies, such as better selecting the predictors, or adding statification or tvc.

That said, please read the FAQ. There you'll find how to share command/data/output as well as pictures of the graphs. This is the best approach to entail insightful replies.

[Soyun Park]

Hello Marcos!
Thank you very much for your advice. I'll definitely have a look at the FAQ for future questions.
I do have some additional questions after trying out the tvc strategy you advised but I'll have to start a different thread for that.

Thank you again!
Best regards,

Soyun