Missing values


Fortran must die
I have a satisfaction analysis where the DV is two levels, the predictors are 4 point likert scale data and some people did not fill out all questions. The default for SAS regression is to remove all individuals who do not answer some predictors (you can change this, but that leads to its own serious dangers based on past review of the literature).

I suggested doing multiple imputations despite having arguably ordinal data, but its been a long time since I did so and I have been told its dangerous/difficult to do this with non-interval data (certainly my own understanding of MI would agree with the difficult part of the coding/analysis). :) We will probably do this eventually but for the present we are only using full responses.

So my question is what are the dangers of using only the responses that have full data (about half the sample answered every question).
@noetsi It seems like you are spreading out your questions over several threads although the seems to be about the same project. That means that we, the readers, lose context. In my view it would be better if you if you kept the different questions in the same thread. (It would be interesting to hear what other "regulars" think of this.)

It seems like you have asked about this project here, here, here and here .

Did you come to a conclusion about your other questions? Was the problem solved?