topic How do I get Hazard Ratio but not using PROC PHREG in Statistical Procedures

How do I get Hazard Ratio but not using PROC PHREG

littleSAScat — Thu, 13 Jul 2017 23:04:10 GMT

Hi All.

I am conducting a survival analysis on a typical clinical dataset in sas studio: Treatment/Placebo; Event/Censored. I am going to use cox model to get a hazard ratio and I know how to do that with PROC PHREG. I am wondering if thats the only way to do it. If it is, how do people validate the value? (From my opinion, validation means different codes lead to same result)

Re: How do I get Hazard Ratio but not using PROC PHREG

Reeza — Thu, 13 Jul 2017 23:31:55 GMT

@littleSAScat wrote:

Hi All.

I am conducting a survival analysis on a typical clinical dataset in sas studio: Treatment/Placebo; Event/Censored. I am going to use cox model to get a hazard ratio and I know how to do that with PROC PHREG. I am wondering if thats the only way to do it. If it is, how do people validate the value? (From my opinion, validation means different codes lead to same result)

I think you're going to need to explain in further detail what you're looking for.

Typically I wouldn't validate the results for every single procedure I ran. But if it was my first time doing survival analysis I'd find some sample examples online or in the documentation and ensure I could replicate the results and that I understood the information. And I'm not sure I would expect 'different codes lead to the same result'. The same result should occur with the same data, but if I've changed my code and therefore my model, why would I expect the same results.

Re: How do I get Hazard Ratio but not using PROC PHREG

Rick_SAS — Mon, 17 Jul 2017 15:33:33 GMT

researchers can use the SAS/IML language to validate or extend the results of SAS procedures. There is a 1994 paper by Ying So that shows how to formulate censored data problems in PROC IML. The appendix has a complete program that you can adapt to your needs.

See also the SAS/IML documentation example "Survival Curve for Interval Censored Data."