Re: Phreg multivariable model - hazard ratio is 1 and significant

SAS_User · Posted 01-21-2017 09:06 PM

I ran a Proc phreg with 2 continuous variables and 3 binary variables. One of the continuous variables (ranges between 10 and 250) is highly significant but the HR is only 1.013. This is from the "Analysis of Maximum Likelihood Estimates" table.

Any thoughts on what might cause such results? I have a reasonable sample size of 150 observations. The binary variables have a distribution of <10 yeses and 140 "no"s

mkeintz · Posted 01-22-2017 12:00 AM

Why is a hazard ratio of 1.013 particularly small for a continuous variable thet ranges from 10 to 250? The HR is presumably for a 1 unit change in the variable (e.g. from 10 to 11). So a 10-unit change has an HR of 1.1378, and the full range of 240 units has an HR of 22.195. Is that disappointingly low?

--------------------------
The hash OUTPUT method will overwrite a SAS data set, but not append. That can be costly. Consider voting for Add a HASH object method which would append a hash object to an existing SAS data set

Would enabling PROC SORT to simultaneously output multiple datasets be useful? Then vote for
Allow PROC SORT to output multiple datasets

--------------------------

SAS_User · Posted 01-22-2017 01:00 AM

Thank you very much for your explanation. My HR is actually exactly 1 for a variable that ranges from .1 to 1600. Would that be just displayed that way due to rounding? How can I explain that? Thanks

SAS_User · Posted 01-22-2017 01:02 AM

It's 1.000 to be exact

JacobSimonsen · Posted 01-22-2017 04:49 AM

I think you should scale your contionous variable such that the variable you put into PHREG goes in from 0 to 10 (or of that order of magnitude).Otherwise rounding errors and errors due to numerical precision otherwise can easily get a too big impact on the interpretation of the estimates.

SAS_User · Posted 02-27-2017 02:56 PM

Thank you for the suggestion. That was very helpful