Is scale(12) exit(120) required?

[Rose Matthews]

Hi

i have a survivalt variable that is already in years specifically 1-17.9 years

i note this code

[code]
stset survivaltime, f(osi==1) scale(12) exit(time 120) //The scale(12) option converts the times recorded in months to years. [code] Source:
https://pclambert.net/software/stpm3/scale_option/

If my time variable - survivaltime is already set into years , as mentioned a range of 1-17.9 years , correct me if I am wrong, but to my mind I don’t need to

scale(12) as its already in years


1. what is the advantage of adding exit(time120) ? // meaning exits at 10 years


2. why include exit(timeX) ?
does it have anything to do with -predict-?

[Clyde Schechter]

You are correct that you do not need the -scale(12)- option when your survival time is already in years.

The -exit()- option, if you use it, tells -stset- that the observation of the study unit ends at the time specified. It simplifies specifying, for example, that all observations that did not fail up to that point are censored at that time. If you have already accounted for that in the failure time variable, then you do not need -exit()- to do it. But if the failure time variable has not been set to that final observation time in all units that have survived to that time, then it is simplest to do it with -exit()-.

The other use of -exit()- is in data that allows the same unit to fail multiple times. In that situation, you must specify -exit()- because otherwise, Stata will assume that any unit that fails exits at the time of its first failure, and so the subsequent failures would not be accounted in the analysis.