Question about the values of the censored column in the OUTCIF= dataset from proc lifetest

[Tiny_Kane]

Hi there, I have the code below to run competing risk analysis and try to plot the CIF curves using proc sgplot. But I found the column of censored in the outcif dataset has values 0, 1, 2, and 3. I guess 0 means event of interest happened, 1 means there was censoring. I wonder what 2 and 3 mean? By the way, my event code is: 0=event free, 1=event of interest, 2=competing events 

proc lifetest data=master2 plots=CIF(test) outcif=cif1 timelist=0 to 60 by 12;
time months_to_scenario1*event_scenario1(0)/eventcode=1;
strata study_id;
run;

data cif1;
set cif1;
if Censored=1 then cenp=cif;
run;
proc sgplot data=cif1;
step x=months_to_scenario1 y=cif/group=study_id;
scatter x=months_to_scenario1 y=cenp / markerattrs=(symbol=plus) GROUP=study_id;
xaxis values=(0 to 60 by 12);
run;

Thank you. 

[FreelanceReinh · Accepted solution]

Hi @Tiny_Kane,

Variable Censored contains the number of censored subjects for the stratum-(group-)time combination summarized by the respective observation of the OUTCIF= dataset. So, value 3, for example, indicates that 3 subjects within the same stratum (and group, if any) must have been censored at the same time. Hence, you should change the conditional definition of variable cenp to, e.g.

if Censored then cenp=cif;

Due to the aggregation to distinct failure times in the OUTCIF= dataset those 3 subjects will be represented by a single plot symbol (even with jittering).

This is really not quite clear from the documentation "OUTCIF= Data Set," which suggests that the original censoring variable is included (which couldn't reflect the aggregation to distinct failure times, though). There are more imperfections: Variable Failcode is not mentioned and variable StdErr is called CIF_STDERR.

The label of variable Censored, "Number Censored," helped me to find this answer (which I confirmed by tests with modified sample data from the documentation).

[Tiny_Kane]

Thank you very much.