I'm analyzing a dataset from a longitudinal study aimed at finding if a set of predictors is associated with the trajectories of an outcome, which is measured each day for seven days. The dataset is in long format and includes the set of predictors (level 2), repeated measures of the outcome and of a potential covariate of the outcome (level 1).
I would like to use a multiple imputation technique which accounts for nested data, preferably with the R package mice. As it is said here: https://stats.stackexchange.com/a/302309, this can be done with 2l. methods. My questions are:
Since I'm dealing with trajectories, should I account somehow for the fact that the outcome measures are "ordered"? Or can I proceed with standard multilevel imputation models?
How can I specify that the imputation process should take into account predictors from both level 1 and level 2 to impute missing values in the outcome variable?