I have data that I think can be best analysed using a survival analysis.

I have bumblebee colonies that need to grow to a certain threshold size for them to become sellable to greenhouse growers. I subjected them to different treatments and want to investigate whether treatment affects the time until they reach this threshold size. Most colonies reach this threshold by the end of the experiment, and some haven't yet but probably will in the future, which should be right censored data.

However, I also have colonies that never grow to the threshold size because an experimental treatment affects them in a way that growth stops, or they show other deleterious traits that cause them to never reach this size/can be sold. This means that the "event" (i.e. reaching the sales size threshold) will never occur for these colonies, and the reason is related to the experimental treatment. How do I deal with these data in a survival analysis? Leaving them out means losing useful information, but leaving them in and censoring them also seems incorrect as this censoring would be "informative", as we know they will never reach the event.

Anyone has ideas how to deal with this?


1 Answer 1


The colonies that will never reach the threshold size can be considered as the colonies for which the event will occur at a very distant point in the future, certainly much longer than the observation time. One could take the time of the event for these colonies equal to infinity - depending on the mathematical/computational expediency.

Stopping of the growth is also an effect of treatment, so generally these colonies have to be kept. However, it also depends on whether such drastic effects are of interest in the experimental context or whether one is interested only in treatments that slightly modify the growth speed.


Your Answer

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

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