So I had a set of 81 items asked in a series of multiple select questions. So for example the first question had 13 and the respondents had to select 4 of those 13. The 31 selections that were made across the questions where then put into a constructed list maxdiff. I don't know which assumption is best to use for the missing values. They are inferior levels, but only to the other items in their question and not to the ones not selected in another question. They are also not missing at random as they are the top items from each question. How would you'll recommend the utilities to be run?