Generalized Estimating Equations (GEE): full factorial or separate interactions?

#1
Hello everyone,

I have a question regarding full-factorial design and separate interactions using Generalized Estimating Equations (GEE). My research is about spatial biases and choice between the left and right sides. By using GEE separate interactions results show that bias is to the left side BUT when I run full factorial design (more interactions automatically come up) I get the results that bias is to the right side.

I am writing my MSc thesis so I need to explain the results.

How should I run the GEE: full factorial or separate interactions? Which and when should I trust? What is the reason of controversial results?

I appreciate your help!