Welcome to the hotseat. We've prepared a guide if you'd like to read more about how it works.

Non-informative censoring - Is it realistic or not?

+1 vote
asked Apr 5, 2017 in BUS 3018F - Models by Maya (430 points)

For question 2.4 in the tut pack, the question reads : " Do you think the assumption of non-informative censoring would be realistic in this case?

I understand that Non-informative censoring is when the censored data gives no information about lifetimes Ti. 
But I am unsure as to how to approach this question. 

1 Answer

0 votes
answered Apr 5, 2017 by asilmotala (2,610 points)
selected Apr 6, 2017 by Rowan
Best answer

Firstly I would just like to emphasise that whenever asking questions, please ensure that you provide all information in the question (don't just reference other sources, as we do not always have those sources on us, as well as it being proper forum etiquette).

So you appear to understand the concept of random censoring, and the idea behind this question is whether it is suitable to apply random censoring to the lives in the prison group (in particular those who left the study early). So as we are looking at mortality, we need to consider if there's any correlation between mortality and length of stay in prison (you could argue this either way - those that remain in prison longer have a higher risk of dying in prison fights or she to poor food or healthcare, or that those that have longer sentences may have worse lifestyles and be riskier, or that those that are being released are simply a random sample). Depending on your argument, you could this either justify the censoring as being appropriate or not. I do not have the memo to this question with me, but my personal opinion is that it is not appropriate, as the risks faced by those who will be leaving soon are significantly different to those who remain in prison, thus not making the censoring truly non-informative.