0

I really confuse about hazard function. Initially, I was thinking hazard function should be a proportion, which indicate if we know some one still alive at time t, so he/she got h(t) (eg. h(t)=20%) chance to die at time t. But through reading, I've been told hazard function is not a proportion, it is a hazard indicator which could >1. If so, the only contribution of hazard model is still to compare individual in different group about the risk at time t. It is seems useless for prediction.

So if I want to predict the chance of die for an individual at a time spot, how can I do it? I'm not interested in which predictor impact the risk either positive/negative or significant/insignificant. I just need to know, with a certain input individual, what is the chance of dying at time t.

Gang
  • 21
  • 2
  • https://stats.stackexchange.com/questions/218947 may also be helpful (these don't explicitly handle the prediction question but the issue here seems to be understanding the concept of hazard function) – Juho Kokkala Jun 28 '17 at 06:48
  • 1
    If time is continuous then the probability of dying at exactly time $t$ (and not some other time, including time $t \pm \varepsilon$ for some very very very small $\varepsilon$), is 0. – Maarten Buis Jun 28 '17 at 07:17

0 Answers0