Confidence interval of median overall survival in Death Watch

[swhjws]

Which analytical method is used for the confidence intervals for the median overall survival in Death Watch? In my case, the CI is not identical with other results which utilized Greenwood formulae.
And please let me know the method is used for the confidence intervals for 1, 3, 5years survival rate.

[jonathon]

hi,

i'm not sure, but you could go noodling through the source code here to find out.

https://github.com/jonathon-love/deathw ... R/surv.b.R

cheers

[patc3]

It looks like it's "log": https://www.rdocumentation.org/packages ... cs/survfit

See code lines 154-170 in the link Jonathon posted, in particular:

Code: Select all

s <- survival::survfit(survival::Surv(times, status) ~ group, data=tmpDat)

This uses the default values from survival::survfit() for confidence intervals, and the documentation states the default method for CIs is log

Also perhaps useful (from the Details section in the documentation):

Based on the work of Link (1984), the log transform is expected to produce the most accurate confidence intervals. If there is heavy censoring, then based on the work of Dorey and Korn (1987) the modified estimate will give a more reliable confidence band for the tails of the curve.

[swhjws]

Thank you to Jonathon and patc3,
Yes, I can find there is not line for conf.type to specify the analysis method. I hope I can use other methods for survival analysis.

Regards,