Assume we have a tire shop and would like to build a survival curve for the lifetime of our tires.Once we know how long our tires are useful, we can send push notifications prompting customers to get their tires replaced.

We know when our tires have been installed, and we have data on change outs with dates. However, some customers buy tires from us then at an unknown time they get their tires replaced at a different store and never do business with us again. So, until we see the customer for tire replacement, we do not know if their last set of tires is still in use or has been changed out elsewhere.

Can we still use records where 'death' is unobserved in this case? How so?

Thank you for giving this the once over fellow gangsters of data.


Unfortunately, I don't think you can. For data to be considered censored, the value must be partially known. For example, imagine a world where people only get there tires changed at your shop. If you have an observation where someone bought a tire from you two years ago, and hasn't been back since, then that is censored data. You don't know how long the tire will last, but you do know that it will be at least two years. You have partial information. But, in your case, once an individual leaves your store, you know nothing. If a person doesnt come back after two years, you don't know that their tires are still working. Sales data for someone who never comes back does not give you any information, and therefore there is nothing useful you can do with it.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.