Should I include all variables in one dataset when imputing missing data using MICE in R?

I have 3 data sets each for a different task with different variables but for the same participants.

I have varying degrees of missing data for each set.
For example in the one task only 70% of participants completed it, whereas in the other task around 10% didn't complete the task (over all each participant completed at least one of the 3 tasks)

My question is should I combine all three data sets into one in order to impute the data using MICE in R? Or is it fine to do separate imputations for each task (data set)?

I have around 40 variables per data set which is why I'm asking before having to sort through that nightmare.


Less is more. Stay pure. Stay poor.
Given your description, it seems like you would merge into one dataset to perform the imputation process on. Then use those imputed sets for the calculation of your test statistic of interest.