If you include zero in the TIMELIST= option in PROC LIFETEST and an event occurs at time zero, the first row in the Product-Limit Survival Estimates table is incorrect. The results indicate that the survival function estimate is 1 and the number of failures is 0.
Here is an example using fictitious data that demonstrates the problem:
data test;
input Days Event;
datalines;
0 1
5 1
9 1
16 0
20 1
20 1
22 0
25 1
;
proc lifetest data=test timelist=0 5 10 15 20 25;
time days*event(0);
run;
input Days Event;
datalines;
0 1
5 1
9 1
16 0
20 1
20 1
22 0
25 1
;
proc lifetest data=test timelist=0 5 10 15 20 25;
time days*event(0);
run;
The Product-Limit Survival Estimates table is:
The correct survival estimate at 0 days for this example is 0.8750. This can be obtained, as well as correct values for the number of failures, standard error, number failed and number left, by omitting the TIMELIST= option.
