I would recommend that you take a step back and think about your problem a bit more. If you have Likert response data as your IVs you may wish to first consider a dimension reduction technique to group them together and turn them into scale variables. This might then lead you to consider a regression model rather than a factorial ANOVA.
By the way, factorial ANOVA will work with Likert response data as your fixed factors. The main problem is small group numbers. The combination of groups suggested by Rana is not compulsory, rather your choice should be led by a desire to make your group sample sizes sufficiently large and approximately equal.