I am struggling a little to know if I am comparing independent or not independent data and I am not sure what kind of test I should use.

Here is the pitch : I would like to compare latencies between two groups of fish:

  • fist group : 14 fish (alone) that had several trials each, so several latencies for one fish
  • second group : 14 different fish that had also several trials, so several latencies for each fish as well. But, those fish are working in pairs with other fish, some coming from the first group, some that are new. Is it enough to talk about non independence between the two groups? And how do I deal with the repeated latencies for one fish? I first thought about paired t test, but it is to compare repeated means of same individuals between groups if I understood well.

Thank you if you can help,


  • $\begingroup$ There's probably not enough information to answer this question. It sounds to me like the design effect will prevent you from actually testing the hypothesis you are interested in: is there a difference in latency between fish of the first group's type and fish of the second group's type. That's because you produce the effect differently. $\endgroup$ – AdamO May 25 '19 at 13:54
  • $\begingroup$ It seems like you might want to use "hierarchical models" - those however may be a bit challenging to get right and it is out of scope to explain them here. There are however some decent tutorials out there. Best of luck! $\endgroup$ – Martin Modrák May 29 '19 at 12:19
  • $\begingroup$ Thank you for your comment, I will either try to find another way to cope with my data or dive into hierarchical models! $\endgroup$ – Morgane la Gaufre May 30 '19 at 14:31

Your Answer

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

Browse other questions tagged or ask your own question.