20 February 2020 2 4K Report

I recently completed an RCT where a portion of the participants were unable to complete some of the outcome measures due to the tasks being too difficult (e.g., at T1, only 90 / 258 kids could complete the flanker).

I have been reading about MAR and MNAR and am unsure which this counts as. My sense is that it's MNAR - kids for whom the task was too hard didn't receive a score - seems very straightforward. However, everything I read about MNAR states that " the probability of being missing varies for reasons that are unknown to us" - and I'd guess maybe some of my data do predict who could and couldn't do the task. For example, age is related to every outcome with older kids doing better, and our study had a wide range of ages (4 - 7). Is it possible these outcome data are then MAR?

Also, if they are MNAR, does anyone have suggestions for what type of sensitivity analyses I should do to back up what tenuous conclusions I can draw from the data? We know we are underpowered and plan to report mostly descriptive results, but will also do some exploratory hypothesis testing and would love to back it up with supplemental analyses if possible.

Thank you!

Similar questions and discussions