Our World Statistics Day conversations have been a great reminder of how much statistics can inform our lives. Do you have an example of how statistics has made a difference in your life? Share your story with the Community!
Choose Language Hide Translation Bar
Highlighted
Level I

Strange inconsistency between JMP(SAS) and R in median survival time estimation

Hi, everyone,

I came across the following strange inconsistency between R and JMP (or SAS) in median survival time estimation.

I understand that CI can be different because of different ways of doing it, but I never expected that median survival point estimation could be different. Does anyone here happen to know the actual technical reason? I tried to figure it out, but had no luck.

For the following simple dataset (event =1 means death, 0 means censor),

 time event trt 116 1 0 116 0 0 110 1 0 116 1 0 116 1 0 110 1 0 116 1 0 116 1 0 150 0 1 150 1 1 138 1 1 138 1 1 150 0 1 150 0 1 150 0 1 138 1 1

JMP cannot provide an estimate of median survival time for trt = 1 group while R can.

JMP output: (JMP 12.1, use analyze --> reliability and survival --> survival, Y = time, grouping = trt, cencor = event, censor code = 0 )

 Group Median Time ... 0 116 ... 1 . ...

R output: (code: library(survival); summary(survfit(Surv(time,event)~trt, data=data))\$table)

 ... median ... trt=0 ... 116 ... trt=1 ... 150 ...

2 REPLIES 2
Highlighted
Staff

Re: Strange inconsistency between JMP(SAS) and R in median survival time estimation

I think what is going on is that there are too many right-censored observations to define the median. For the 8 observations in the trt=1 group, 3 are failures at t=138, 1 is a failure at t=150, and half (4) of them are right-censored at t=150. Since the median of 8 values is the average of the 4th and 5th values, we would need to average 150 and some (unknown) value larger than 150. As a result, the Survival platform is reporting a missing value for the Median Time.

As to why R gives a value, I can only speculate, but my guess is that since the 4th ordered value is not censored and the 4th observation out of 8 would be the 0.50 quantile, R is reporting the value of the 4th observation as the median survival time. Note that if you change one more t=150 observation (row 10) to be right-censored, R reports the median for trt=1 to be missing (NA).

Hope this helps,
Michael

Michael Crotty
Sr Statistical Writer
JMP Development
Highlighted
Level I

Re: Strange inconsistency between JMP(SAS) and R in median survival time estimation

Thanks a lot for the explanation! That makes sense!

Now I have to see if I can find a formula or something that gives a clear definition in SAS vs R, since I need to let R output the SAS result.

Thanks again!

Best,

Yan

Article Labels

There are no labels assigned to this post.