Prroc phreg output statement

[Yasu]

Hi,

I'm applying two different cox models to the same dataset.

The linear predictor (XBETA) is different, but the survivor function estimate is exactly the same between two models.

Does this happen by chance?

Below are codes and outputs:

PROC PHREG DATA=&DATA.;
CLASS Z/PARAM=REF REF=FIRST;
MODEL TIME*EVENT(0)=Z P1_3 P2_3 / RL;
OUTPUT OUT=A SURVIVAL=SURV XBETA=XBETA;
RUN;

PROC PHREG DATA=&DATA.;
CLASS Z/PARAM=REF REF=FIRST;
MODEL TIME*EVENT(0)=Z SP1_3 SP2_3 / RL;
OUTPUT OUT=B SURVIVAL=SURV XBETA=XBETA;
RUN;

[PaigeMiller]

So the difference in the two models is the two different model statements:

MODEL TIME*EVENT(0)=Z P1_3 P2_3 / RL;
MODEL TIME*EVENT(0)=Z SP1_3 SP2_3 / RL;

Since you know, and we don't, what is different between P1_3 and SP1_3 and what is different between P2_3 and SP2_3, perhaps you could tell us what the differences are. In particular, if there is a mathematical function that gives an exact relationship between P2_3 and SP2_3 and an exact relationship between P1_3 and SP1_3, then the fact that the survivor function estimates are identical would be expected.

[Yasu]

Hi Paige,

Thank you for your reply.

They are imaginary variables that are normally distributed. Please imagine that P1_3 and P2_3 are office systolic and diastolic blood pressure, and SP1_3 and SP2_3 are home systolic and diastolic blood pressure. 

Thanks,

Yasu

Yasu

[PaigeMiller]

Were they created to have some mathematical relationship between the office variable and the home variable?

[Yasu]

They were simulated so that they have weak correlations among them.