R 与 SPSS 结果的置信区间不同

Question

I have a small dataset with the total number of cases under 20. When I ran survival analysis using Kaplan-Meier estimate on SPSS with this dataset, I was able to get specific 95% upper and lower confidence interval values, but in R (using survfit, conf.type "log-log") 0.95UCL is not estimated. I get the feeling that this happens often when I get to use small datasets. I'm not sure if it's accurate but I've read that SPSS provides log-log as default. So, could anyone let me know why the result is like this? Thank you in advance.

Answer 1

I believe the confidence intervals in SPSS KM are equivalent to the "plain" option in R's survfit, not the "log-log" option. You can see the formulas in the KM chapter of the IBM SPSS Statistics Algorithms manual .