Default reference group in SAS proc

[Cruise]

Hi Folks:

In PROC PHREG below where METHOD is binary variable (1,0) does SAS take 1 over 0 or 0 over 1 to calculate the hazard ratio reported in the annotated output shown below? 

PROC PHREG DATA=MYDATA; 
MODEL DURATION*DEATH(0)=METHOD;      
RUN;

[FreelanceReinh · Accepted solution]

Hi @Cruise,

Without a CLASS statement, METHOD is treated as a continuous variable, in which case the hazard ratio refers to an increase by one unit. For a binary variable, of course, this is equivalent to a hazard ratio of level 1 vs. level 0. Categorical explanatory variables are usually named in a CLASS statement where you can use the REF= option to specify the reference level explicitly and thus avoid any ambiguities.

Your sample output shows one of the rare cases where the "direction" (1 vs. 0 or 0 vs. 1) has no impact on the results (hazard ratio 1).

[Cruise]

Can you please elaborate what made the case one of the rare cases where the "direction" (1 vs. 0 or 0 vs. 1) ?

[FreelanceReinh]

When you switch the two levels of a dichotomous explanatory variable (used in a CLASS statement), the original hazard ratio h becomes 1/h. Similarly, the confidence limits (CL) of h are inverted: The new lower CL is 1/(old upper CL) and the new upper CL is 1/(old lower CL). Also, the sign of the parameter estimate, if positive or negative, changes (but not its absolute value, standard error, Chi-Square statistic and p-value).

However, in your example h=1 (as a consequence, the upper CL is 1/(lower CL)) and the parameter estimate is 0 so that none of the changes described above is visible.

[Cruise]

Thanks for your explanation. This helped!

[Reeza]

I always recommend using HazardRatio statements to interpret your HR so that you know exactly what comparisons it's making.