I am investigating the relationship between pornography consumption and several measures of sexism. Responses were given on likert scales. Since participants were able to leave out questions in the survey, missing values occured (less than 5 % per item).

I would like to replace these values to not lose 20 cases because of missing data on single items. I tried to use the estimation maximization (EM) function in SPSS, but it produced values outside the scales (e.g. likert scale 1-7, estimation: 0, -1).

Is there a method that is recommendable for my data preferably in SPSS?

PS: I don't have much experience with inputation.

Thank very much you in advance!