I wondered if you could help me.
I have run a multinomial logistic regression with a DV with three levels (young, middle aged, old) and three IV's (three questionnaire items). I have found two significant main effects, a significant two-way interaction and a significant three-way interaction for young relative to old. I only found one significant main effect and a significant three-way interaction for middle age relative to old. However, the graphs that I have produced show that it is middle age that has a completely different trend of results from both young and old, who have very similar results.
I am really confused as to what these significant interactions mean? Are they significant relative to only old or do they reflect all three categories, meaning that young relative to old has a significant two-way interaction because they both differ from middle age?
Can anyone shed any light?