0
$\begingroup$

I have a question related to survival models: for instance, the survival probability of people that have a disease.

In all examples I've seen so far, the data used to estimate those models sometimes includes censored observations (people who are sick but have not died yet) but always includes non-censored observations (people who died and the life duration of whom we therefore know).

Now suppose you only have the sub-sample of censored observations (i.e people who are still alive, no one in the sample has died yet). I'm guessing a survival model will give biased estimates (lower than the actual survival), but that models like simple OLS are also biased... So I was wondering which model would be more appropriate for that type of data?

Thanks a lot!

Edit: There are averages for the whole population available online, for that variable (how long people stayed in their house). But at individual level, which is the data I want to use, that variable is not available, and all I have is when the individuals moved into their house (and so how long they stayed until now, but they might stay a few more years).

The question I want to answer is, given some individual characteristics, how to predict how long people stay in a house.

I hope this is clearer, thanks again!

$\endgroup$
  • $\begingroup$ It's impossible to tell you what model would be appropriate without knowing what question you are trying to ask of the data. What are you hypothesizing is the case with this sub-sample of censored observations, and why are you looking at this sub-sample instead of the whole sample? $\endgroup$ – Ryan Simmons Jul 17 '18 at 13:25
  • $\begingroup$ Thank you for your answer! I used the mortality example because it's the most common, but my data is actually real estate data, where we know when people moved in their current house, but what we want to know is how long they stay in this house before moving in a new one. We only have averages on that second variable, so there is no other possibility than to work on this 'sub'sample of people who have not yet left their house. $\endgroup$ – AlC Jul 17 '18 at 13:38
  • $\begingroup$ I'm sorry, but I'm still not entirely sure I understand. In particular, the last sentence of your comment I have read 10 times and still don't quite get what you mean. What, exactly, do you have averages of, and why does having averages lead you to believe there is "no other possibility than to work on this 'sub' sample"? Could you more precisely describe exactly what data you have, and what question you are trying to answer with it (preferably in an edit to the OP, rather than in the comments)? $\endgroup$ – Ryan Simmons Jul 17 '18 at 19:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.