I read a bunch of references on the bias that one should beware of when analyzing survival times. One classic example is the one of Oscar winners: a miss-specification of the model leads to two kind of bias, the length bias and the time-dependent bias. See the figure below. enter image description here

a) correct model;

b) length bias;

c) time-dependent bias;

d) both length and time-dependent bias.

However, once I have specified the right model, how can I compare the two groups (oscar winners vs others)?

  • $\begingroup$ How do you define 'not Oscar winners', for example given not all Oscar winners are born in the same country. If you define the problem along the lines of 'does the average American born Oscar winner live longer than the average American', it seems at least possible that they do, given movie stars are usually wealthy, hence with better access to health care than people in general. Also, by defintion, they don't work in a dangerous occupation comparable with say 'soldier' or 'emergency services worker'. (you could use life tables to avoid the time dependent problem in (b) ) $\endgroup$ May 15, 2017 at 11:21


Your Answer

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

Browse other questions tagged or ask your own question.