Thank you so much, I am not perfect in statistic and SEM, I am just a beginner. Your advice was useful to me.
Your link that you sent me was helpful. I ran SEM without residuals because I am not sure if I can use them in the model based on my purpose. I think that I need to learn more about residuals. Of course, I found some papers that they used the residuals in the model.
I just want to that the basis for Phil's solution is that a linear model can simultaneously account for all the contributions to each response. You designed a multiple factor experiment but you are analyzing it one factor at a time. It is not an invalid approach, but you can do more with the model that Phil proposed. For example, you can use the Prediction Profiler to visualize the dependence on all the factors.