As a hobby application, I want to estimate a Kaplan-Meier curve for "survival" of football referees, i.e., for how long they stick to this activity. I have data for when they started and, if applicable, when they quit, or alternatively, that they are still active.

Now, the problem is that digital records only started about 15 years ago. So if a referee quit and/or has started during the last 15 years or has been active for more than 15 years, I know that. What I do not see is referees who started, say, in 1990 and quit in 1998. Given the problem of many referees quitting early, this seems to introduce a bias due to missing data.

I could use only data for referees who started refereeing during the last 15 years, which would however throw away much data on long-serving referees.

Is there any appropriate way to deal with this situation?

  • $\begingroup$ is this called left trucated? $\endgroup$
    – Deep North
    Sep 26, 2017 at 11:11
  • $\begingroup$ I am not sure, this is my first foray ever into survival analysis...I would guess no, however, as I do not see these 20th century quitter observations at all, while I am sure there will have been such early quitters did exist before the digital records started. $\endgroup$ Sep 26, 2017 at 11:16
  • $\begingroup$ I think left truncated means they have been there when the sample started, but one does not know for how long they had been there before the sample started. In my case, if I see an entry, I do know when he/she started refereeing. $\endgroup$ Sep 26, 2017 at 11:28


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Browse other questions tagged or ask your own question.