I have a student who has collected observation data of distraction behaviours (using a checklist of behaviours) in autistic children under three music/sound conditions - this is repeated measures. She wants to compare total distraction behaviours amongst the three conditions in general. Initially I thought Chi-squared, but as it's repeated measures, thought McNemar, but then 3 conditions so now I don't know.... which test does this need?
However, then she also wants to see whether different 'types' of distraction behaviours are affected differently by the sound conditions. The checklist has been divided into categories based on 'types' of distraction. She would therefore be looking to run the above test on each of these categories to determine how they are each affected by the sound conditions, and then compare results of tests between these different 'types' of distraction to test for significant differences. Is there a test that can do this comparison?
TIA.
This is, I believe, a set of complicated longitudinal models.
https://www.stat.ncsu.edu/people/davidian/courses/st790/notes.html
This link is to the best set of notes I know of on these models. Hopefully this will help. Best regards, David
Sounds like a multilevel Poisson regression is appropriate, but that assumes that your counts aren’t regularly running into the ceiling. If so, it’s not really a Poisson. If so, it might be best to do a multilevel logistic with behavior treated as an item random effect. Apologies if you don’t know multilevel modeling.
Well the rhs of the model is longitudinal. If the counts on the lhs require poisson regression you may be in serious research areas . I haven't heard of anything like that. I don't know how to do an estimation like that. Is there a transformation that might help???? I guess this is why we tell people to lock down the analysis prior to collecting data. Sorry. David
A long shot. Any chance something like chapter 7 of the above notes by Davidian would be helpful? They do cover some non-continuos situations. And chapter 8 goes into how to apply chapter 7 ideas to particular situations. To be honest this looks like a good dissertation to me unless someone has happened to already do it. Sorry. BTW Davidian has a couple of advanced books in this area. Best wishes.
Thank you so much for your answers. This has escalated quickly into territory I'm not very familiar with, but thank you for pointing me in the right direction. Yes, I definitely should have considered the analysis in more depth at the outset... lesson learned.