I am attempting to analyze data for someone and am getting stuck on finding something that is appropriate. The design of the experiment is as follows: vials are set up with 10 flies per vial (either mutant or control flies). Flies are tapped to the bottom of the vial and the time it takes half the flies to reach a certain height is recorded. Unfortunately, the observer watched for 90 seconds and if half the flies had not reached the designated height by this time the data was recorded as >90. My first thought was to use survival analysis since it can handle censoring. However, if I understand correctly, these right censored data would be considered informative, since flies that did not reach the height by 90 seconds are obviously slower, and might, in fact, never reach the designated height. This could also correlate with the level of the predictor (mutant vs. control). From my understanding a standard survival analysis would not be appropriate here. I have two questions:

  1. Is my understanding of informative right censoring correct?
  2. If so, does anyone have any suggestions for how I could analyze this dataset? Obviously one solution would be to have the observer continue watching for longer, but, aside from the fact that much of the data has already been collected, I'm also not sure it would solve the problem because the flies may be so impaired that they never reach the designated point. Thank you for reading and I appreciate any input!

1 Answer 1

  1. No. The condition for noninformative right censoring is that, among vials that reach 90s without half the flies getting to the desired height, whether observation continues or stops is independent of how much longer the flies would take. This is true: according to what you say, observation always stops at 90s

  2. Survival analysis is one sensible approach.


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.