Cox regression analysis when main exposure is a three-level categorical variable

[_MooMoo]

Hello SAS communities! 

I am working with time-to-event data and the main exposure is a three level categorical variable (X is either 0 or 1 or 2). I ran proc phreg and got two hazard ratio estimates which compares 1) 0 vs. 1 2) 0 vs. 2 when I am setting X=0 as reference level. I am wondering if this statistically means the same thing when I subset two data sets, one with X=0 and 1 and another with X=0 and 2. And if I run two separate cox regression, will I get the same estimate? 

Any advice is appreciated!

[OsoGris · Accepted solution]

I don't think you will get the same estimates if you split the data.  It makes more sense to use one data set for this analysis.