I have an advanced cancer quality of life dataset, where missing data is not only non-ignorable, but also non-monotone as at some time points subjects missed filling out a question but didn't drop out of the study. Is there any literature to describe how to model such missing data that is non-monotone in MPlus? Thank you!
Both monotone and non-monotone missing data can be handled in Mplus with ML under the standard MAR assumption. Non-ignorable missing data, that is non-MAR (NMAR), is more typically modeled to account for dropout, that is monotone missingness. The assumption probably is that non-monotone missingness is more likely to be MAR and NMAR methods are mostly needed for monotone situations.
I discussed NMAR modeling in
Muthén, B., Asparouhov, T., Hunter, A. & Leuchter, A. (2011). Growth modeling with non-ignorable dropout: Alternative analyses of the STAR*D antidepressant trial. Psychological Methods, 16, 17-33.
These models also take care of non-monotone missingness that is MAR.
NMAR modeling for non-monotone missingness is also possible, say using pattern-mixture techniques (see paper).