## End of Final Interval with Proc Lifetest

Frequent Contributor
Posts: 80

# End of Final Interval with Proc Lifetest

Please ignore. I figured out an issue I had with data coding. (Those surviving to year 1 are all censored in the interval [12, .} whereas I was coding them to be censored in [11, 12)) Sorry!

I am using Proc Lifetest to compute survival estimates for interval censored data, where the interval is 1 month (30 days) and the study is 1 year. My goal is to state and show graphically the probability of surviving to year 1. Here is my code:

proc lifeTest data = lifeData method = lt intervals =(0 to 12 by 1);
time Month*Censored(1);
run;

My problem is that this code only allows me to state and show the probability of surviving to (the beginning of) month 12, not to the end of month 12 (1 year). The data needed to compute the 1-year survival probability exists in the last row of the existing SAS output; LIFETEST just needs to go an extra step to produce one more row (of course, most values will be missing).

Ex.: The last two rows of my results include:

Interval [10, 11) Failure = 0.00360 Cuml. Survival = 0.9757
Interval [11, 12) Failure = 0.00743 Cuml. Survival = 0.9722 [= 0.9757*(1-0.00360)]

I would like a final row that says:

Interval [12, .) Failure = . Cuml. Survival = 0.9650 [= 0.9722*(1-0.00743)]

I realize the numbers are easy to code with a data step afterwards, but I need the graph to include this final value with confidence limits as well: PLOTS = (SURVIVAL (CL)). I tried incrementing my intervals to 13, but it does not work. Any thoughts?

Thanks!

Edit: Because my total N is large, I can get the correct values by adding in a dummy record censored in [12, 13); however, my effective sample size is then 1 too large for each row. That may work, but doesn't seem like an very elegant solution.

Message was edited by: Kastchei Message was edited by: Kastchei
Discussion stats
• 0 replies
• 168 views
• 0 likes
• 1 in conversation