Estimating a survival probability in R (2)

Asked Apr 19 '12 at 07:10

Active Apr 19 '12 at 07:10

Viewed 820 times

Following Brian Diggs's answer, I have tried to estimate some survival probability in R. However, the result is wrong (1 instead of 0.9) and I do not understand what is going on... Can you explain me?

> time
 [1] 1.38103330 2.14191380 1.47271212 0.02703924 0.34921793 3.01005383 1.01132532 0.09546903 0.44729971 1.12909544
> event
 [1] 1 1 1 1 1 1 1 1 1 1
> km <- survfit(Surv(time, event) ~ 1)
> surv <- stepfun(km$time, c(1, km$surv), right=FALSE, f=0)
> surv(min(time))
[1] 1

edited Apr 13 '17 at 12:44

Community

asked Apr 19 '12 at 07:10

user7064

1,685
5
23
39

1

Here you use the left-continuous version of the Kaplan-Meier estimate (option right=FALSE), so the result is correct. – Stéphane Laurent Apr 19 '12 at 07:14
@Stéphane Laurent: But according to the 'help' page, "For the default, right = FALSE, f = 0, fn is a cadlag function, i.e., continuous at right, limit (‘the point’) at left.". – user7064 Apr 19 '12 at 07:20
Strange - you could also try approxfun() with option method="constant" – Stéphane Laurent Apr 19 '12 at 07:35
Can I have an example with approxfun() ? – user7064 Apr 19 '12 at 08:09
1

Try this: surv – Stéphane Laurent Apr 19 '12 at 08:19
Moreover I remember I already had a similar problem, but I do not remember well... instead of km$time, try to use sort(time), does it work ? – Stéphane Laurent Apr 19 '12 at 08:50
1

It looks to work with 'sort(time)'. This is a bit wierd... but many thanks! – user7064 Apr 19 '12 at 09:02
1

In passing do not use "time" as a variable name ! This is the name of a R function and this can be dangerous ! – Stéphane Laurent Apr 19 '12 at 09:05
1

Ok... thanks. By the way, the solution with "sort(time)" may fail in case of ties... – user7064 Apr 19 '12 at 09:08
1

Then type sort(unique(time)) – Stéphane Laurent Apr 19 '12 at 09:22

Estimating a survival probability in R (2)

0 Answers0