Hi,

I am trying to do multiple imputation with PMM using IBM SPSS version 27. I have 70 variables in my imputation model, and nearly 22 thousand cases. Thus, accounting for twoway interactions is not feasible (at least with my computer). But, what level of information is 'lost' if I have INTERACTIONS NONE? After imputation I am to do regression analysis and compare means.

The second issue is related to imputation of implausible values. I have a questionnaire that consists of 10 questions (coded as individual ordinal variables). But if one has answered the first question 'never' then the latter ones cannot have answers like 'once a week' or so, but should also be 'never'. However after multiple imputation, I see (did a cross tabulation with first question vs other nine questions), that there are plenty of these kind of implausible values imputed. I have already made changes in my imputation model: I tried change the order of the variables and also tried those interactions in a sub-sample of variables, but none of those have helped in resolving this issue. Therefore, I was wondering if it is possible to code restrictions in SPSS Syntax? The restriction would be that the imputed value for q1 is the biggest/maximum value of the q2-q10 or in other words that it's value cannot be smaller than that with biggest value of the latter once.

the MI procedure does not support constraints such as you specified. You might want to just override the imputed values with COMPUTEs for that situation.