A longitudinal dataset has some missed visits. The probability of missing a visit is related to participants' previously observed stress scores, and those scores are included in the analysis model. Which assumption may support likelihood-based estimation under this pattern?