Created:
| Last Modified:
Hi,
Why the upper 95% value not calculated for my survival analysis data?
Thanks
Neeraj
1 ACCEPTED SOLUTION
Accepted Solutions
Created:
| Last Modified:
It is possible that the upper confidence interval is missing. Actually, both ends can be missing. Or even the median estimate itself can be missing. All depend on your data.
To understand the reason, I will try something plain, but the wording may be less rigorous.
Suppose we have three distinct event times at 1, 2, and 3. First two are failures, and the last one is a censored observation.
And here is the report.
The estimate of the median is time = 2, at which the estimate of the survival probability is 0.3333, which is less than 0.6667 (larger than 0.5) at time = 1. So the upper end of the confidence interval should be some time point after 2 and at which the survival probability estimate is even less and satisfies the condition described by the paper below. However, there does not exist even a single such time point after time = 2. Time = 2 is the last failed observation
The rigorous explanation can be found in this paper:
Brookmeyer, R., & Crowley, J. (1982). A Confidence Interval for the Median Survival Time. Biometrics, 38(1), 29-41. doi:10.2307/2530286
Created:
| Last Modified:
It is possible that the upper confidence interval is missing. Actually, both ends can be missing. Or even the median estimate itself can be missing. All depend on your data.
To understand the reason, I will try something plain, but the wording may be less rigorous.
Suppose we have three distinct event times at 1, 2, and 3. First two are failures, and the last one is a censored observation.
And here is the report.
The estimate of the median is time = 2, at which the estimate of the survival probability is 0.3333, which is less than 0.6667 (larger than 0.5) at time = 1. So the upper end of the confidence interval should be some time point after 2 and at which the survival probability estimate is even less and satisfies the condition described by the paper below. However, there does not exist even a single such time point after time = 2. Time = 2 is the last failed observation
The rigorous explanation can be found in this paper:
Brookmeyer, R., & Crowley, J. (1982). A Confidence Interval for the Median Survival Time. Biometrics, 38(1), 29-41. doi:10.2307/2530286